Genotype-Dependent Dual Effects of Zinc on Cadmium Accumulation in Rice Revealed by a Multi-Scale Quantitative Framework

Genotype-Dependent Dual Effects of Zinc on Cadmium Accumulation in Rice Revealed by a Multi-Scale Quantitative Framework | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Genotype-Dependent Dual Effects of Zinc on Cadmium Accumulation in Rice Revealed by a Multi-Scale Quantitative Framework Shimiao Chen, Bin Shan, Fuhai Zheng, Yanyan Li, Xi Chen, Qinyu Lu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9361934/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Cadmium (Cd) contamination poses a major risk to rice safety, while zinc (Zn) can modify Cd accumulation in a genotype-dependent manner, with its biological basis remaining incompletely understood. Here, 44 rice varieties were hydroponically cultured under Cd stress either alone or under a near-equimolar Zn + Cd co-exposure treatment, and Zn–Cd interaction patterns were analyzed using an integrated framework combining response-landscape quantification, Random Forest modeling, and sparse partial least squares (sPLS) analysis. The response landscape revealed a four-quadrant distribution of interaction types across cultivars, with response intensities spanning nearly a 200-fold range, indicating that Zn effects on Cd accumulation were highly variable rather than uniformly inhibitory. Random Forest classification distinguished response direction with ~ 70% accuracy using baseline metal status and Zn-induced physiological shifts as predictors, suggesting that non-random information about response direction is recoverable from surface-level phenotypic variation. sPLS analysis of 12 representative cultivars further resolved layer-specific candidate signal structures, with translocation-related variation associated primarily with a root superoxide/flavonoid-centered structure, leaf Cd accumulation with a more flavonoid/hormone-centered structure, and total Cd amplitude with a phenolic/flavonoid-associated structure. Together, these results show that Zn effects on Cd accumulation are genotype-dependent and identify the root redox baseline state under Cd stress alone as a candidate upstream feature associated with response direction. They further suggest that contrasting outcomes across cultivars may reflect different state-dependent response regimes linked to baseline redox status. This framework provides a basis for understanding why the same Zn intervention can yield opposite Cd outcomes across rice cultivars, and it requires broader validation before any screening or management application is proposed. Rice Cadmium accumulation Zinc–cadium interaction Genotype-dependent response Random Forest Sparse partial least squares Figures Figure 1 Figure 2 Figure 3 Introduction Zinc (Zn) and cadmium (Cd) co-occur widely in agricultural soils, particularly in paddy systems affected by phosphate fertilizer inputs, mining activities, or industrial emissions (Hussain et al., 2021 ). Because Zn²⁺ and Cd²⁺ share similar ionic radii and charge properties, they compete for common uptake and translocation pathways, including ZIP-family transporters and low-affinity cation channels such as calcium-permeable channels (Cai et al., 2019 ; Hu, 2021 ). This structural and chemical similarity means that elevated Zn availability can saturate shared transport systems, effectively reducing Cd influx into root cells—the mechanistic basis for antagonistic Zn–Cd interactions (Wang et al., 2018 ). Under such conditions, Zn supplementation, including foliar application of ZnSO₄, has been shown to reduce Cd concentrations in plant tissues and grain substantially, and to enhance antioxidant enzyme activities that mitigate Cd-induced oxidative stress (Rizwan et al., 2019 ; Wang et al., 2018 ). Beyond competitive inhibition at the transporter level, Zn may also influence rhizosphere chemistry, altering soil pH and organic matter interactions and thereby modifying Cd speciation and bioavailability before root uptake (Du et al., 2020 ). However, the outcome of Zn–Cd co-exposure is not uniformly antagonistic, and its direction is strongly dependent on the relative concentrations and molar ratios of the two metals, as well as on soil chemical properties and plant physiological status (Cai et al., 2019 ). Under Zn-deficient conditions, reduced competitive inhibition can paradoxically enhance Cd uptake and translocation, yielding synergistic interactions and elevated Cd burden in plant tissues (Haider et al., 2021 ). Even under adequate Zn supply, certain cultivar-specific responses defy the simple antagonism model: Zhou et al. observed that simultaneous Cd–Zn contamination suppressed Zn uptake while paradoxically elevating Cd accumulation in rice tissues, illustrating that the directionality of the interaction cannot be predicted from metal concentrations alone (Shahzad et al., 2025 ). Zn supplementation may further modify the bioaccessibility and gastrointestinal solubility of grain Cd — through alterations in Cd speciation and binding within grain tissues — thereby altering actual human exposure risk in ways that total Cd concentration alone would fail to capture (Vance and Chun, 2015 ; Wang et al., 2024 ). Collectively, these bidirectional, ratio-dependent, and context-sensitive outcomes reflect an interaction system whose direction and magnitude are jointly governed by soil chemistry, metal speciation, plant physiology, and cultivation context (Du et al., 2020 ; Qixing et al., 1994 ). The complexity of Zn–Cd interactions is further shaped by the substantial genotypic variation that exists among rice cultivars in their capacity to take up, translocate, and sequester Cd. Rice ( Oryza sativa L.) is the primary dietary Cd exposure route for over half the global population (Vu et al., 2022 ), and its efficient root-to-grain Cd translocation makes it a crop of particular concern under contaminated field conditions (Zhang et al., 2021 ). Cd-tolerant genotypes typically exhibit more robust antioxidant defense systems, more efficient Cd immobilization in the cell wall fraction, and greater vacuolar sequestration capacity, collectively reducing Cd translocation to aboveground tissues and limiting its deposition in grain (Chang et al., 2023 ). Comparative metabolomic profiling of contrasting indica varieties — the Cd-tolerant NH224 and the Cd-sensitive NH199 — demonstrated that tolerance was associated with enhanced amino acid biosynthesis, elevated hormone metabolism, strengthened phenylpropanoid pathway activity, and upregulated antioxidant enzyme systems, illustrating the breadth of metabolic reprogramming that underlies genotypic Cd resilience (Chang et al., 2023 ). At the molecular level, this variation is mechanistically anchored in the differential expression and functional activity of a small set of membrane transporters. OsNramp5, localized to root epidermal cells, mediates initial Cd entry from the rhizosphere and represents the primary influx route; OsHMA2 , expressed in root vascular tissue, controls xylem loading and root-to-shoot translocation; and OsHMA3 , a tonoplast-localized vacuolar transporter highly expressed in root cells, sequesters Cd into vacuoles and thereby restricts its upward movement toward aerial tissues and grain (Ueno et al., 2010 ). Allelic polymorphisms and genotype-specific transcriptional regulation of these transporters — modulated by Cd-responsive signaling pathways, phytohormone fluctuations, and oxidative stress cues — drive marked cultivar-level differences in Cd partitioning (Sasaki et al., 2014 ; Wang et al., 2020 ). Cultivars harboring functional OsHMA3 alleles exhibit substantially lower shoot and grain Cd accumulation compared to those lacking them, underscoring the central role of vacuolar sequestration capacity as a determinant of grain safety (Sasaki et al., 2014 ). Critically, because these same transporter systems mediate both baseline Cd handling and the competitive dynamics between Zn²⁺ and Cd²⁺ at shared binding sites, genotypic background is likely a primary determinant of how a given cultivar responds to Zn co-exposure — yet this connection has rarely been examined systematically (Adil et al., 2020 ; Tavarez et al., 2023 ). Despite this mechanistic foundation, understanding of how Zn–Cd interactions manifest across cultivated rice germplasm remains limited and fragmented. Existing studies have reported both antagonistic and synergistic outcomes. Still, these observations are usually confined to individual cultivars or narrow experimental settings, and the reported directions of interaction often contradict one another across studies (Tavarez et al., 2022 ). Some studies describe robust suppression of grain Cd following Zn supplementation. In contrast, others report paradoxical increases in grain Cd in particular genotypes, with these discrepancies attributed to differences in root exudation, rhizosphere pH, transporter affinity, or cultivar-specific metal partitioning (Fontanili et al., 2016 ; Tavarez et al., 2023 ). Interpretation is further complicated by substantial methodological heterogeneity, including differences in Zn and Cd concentrations, Zn/Cd molar ratios, treatment timing, and the tissues selected for metal quantification, all of which make direct cross-study comparison difficult (Shahzad et al., 2025 ). Crucially, no study to date has systematically characterized the full spectrum of Zn–Cd interaction outcomes — from strong antagonism to strong synergism — across a representative cultivar panel under controlled and directly comparable exposure conditions (Shahzad et al., 2025 ; Tavarez et al., 2022 ). As a result, it remains unclear whether the contradictory findings in the literature primarily reflect genuine biological diversity among genotypes, methodological heterogeneity, or both. Equally lacking is a structured framework for identifying the cultivar-level biological features associated with whether Zn supplementation inhibits or promotes Cd accumulation, and for determining whether response direction and response magnitude are governed by the same or distinct biological axes. Two specific and interrelated questions, therefore, remain unresolved. First, which cultivar-level biological features - encompassing baseline metal accumulation status, Zn-induced shifts in transporter expression, redox and antioxidant responses, hormonal signaling, and secondary-metabolite mobilization - are associated with whether a genotype responds to Zn co-exposure with net Cd inhibition or net Cd enhancement (Shahzad et al., 2025 )? If response direction is associated with measurable cultivar traits, it becomes possible to stratify germplasm by likely interaction pattern under defined co-exposure conditions and to assess whether cultivar background should be considered when interpreting Zn-based mitigation outcomes. More broadly, this also raises the question of how reliably Zn-based mitigation effects can be generalized across cultivar backgrounds, because conclusions drawn from single-cultivar or narrow-panel studies may not fully capture the diversity of outcomes observed in broader rice germplasm (Shahzad et al., 2025 ; Tavarez et al., 2022 ). Second, are the biological features associated with response direction the same as those that govern response magnitude, or do these represent partially distinct biological axes? This distinction matters because direction and amplitude may reflect different physiological processes operating at different regulatory levels. Disentangling these two dimensions requires an analytical framework capable of handling multivariate, multi-layered biological signals simultaneously rather than examining single variables or pathways in isolation (Mashabela et al., 2023 ). Resolving these questions is important both for building a mechanistic understanding of cultivar-specific Zn-Cd interaction and for defining testable candidate features that may condition interaction outcome under combined exposure (Hussain et al., 2021 ; Zhang et al., 2021 ). To address these gaps, the present study investigated Zn-Cd interaction patterns across 44 rice cultivars under a fixed near-equimolar co-exposure regime. This design was chosen not to characterize the full agronomic dose-dependence of Zn supplementation, but to enable a controlled, directly comparable assessment of cultivar-specific interaction outcomes under equivalent molar inputs of the two metals. An integrated three-tier analytical framework was employed. A quantitative response landscape was first constructed to characterize the directional and magnitude diversity of Zn-Cd interactions at the population level across all 44 cultivars. Random Forest modeling was then applied to identify the cultivar-level features - comprising baseline metal status and Zn-induced physiological shifts - that were most informative for response direction and magnitude. Finally, sparse partial least squares analysis was performed on a subset of 12 representative cultivars using a mechanistically enriched variable set to identify candidate signal structures associated with distinct Zn-Cd response layers, including transporter expression, ROS/antioxidant status, hormonal signaling, and secondary metabolite profiles (Chun and Keleş, 2010 ). This multi-scale approach was designed to characterize population-level diversity in Zn–Cd interaction outcomes and to define candidate biological features associated with response direction and magnitude under the tested co-exposure condition, while also providing a broader framework for understanding cultivar-specific differences in constitutive Cd accumulation tendency and co-exposure responsiveness. Materials and Methods Plant materials, treatments, and sampling Forty-four rice varieties ( Oryza sativa L.), comprising Indica and Japonica genotypes, were sourced from local germplasm collections and commercially cultivated varieties widely planted across Guangxi, China (Table S1 ). Seeds were surface-sterilized with 2.5% sodium hypochlorite for 10 min and rinsed thoroughly with distilled water before germination. Seeds were germinated in the dark at 28°C for 2 days, then sown onto the seedling substrate. Uniform seedlings were transferred to a half-strength Yoshida nutrient solution (1971) for hydroponic cultivation at the one-leaf-one-heart stage. After reaching the three-leaf–one-heart stage (~ 7 days), Cd and Zn treatments were applied. The nutrient solution was aerated continuously, refreshed every 3 days, and maintained at pH 5.5 ± 0.1. The samples were then divided into two groups. The control group was grown in full-strength Yoshida nutrient solution containing 0.2 mg/L Cd alone. In comparison, the treatment group was simultaneously exposed to 0.2 mg/L Cd and 0.12 mg/L Zn for 14 days under controlled hydroponic conditions. The Cd concentration (0.2 mg/L) was selected based on previously reported contamination-relevant exposure levels(Wen et al., 2024 ). The Zn concentration (0.12 mg/L) was chosen to provide a near-equimolar co-exposure relative to Cd, so that the two metals were supplied at comparable molar intensity within a fixed dual-metal regime. This design was intended to facilitate interpretation of competitive and genotype-dependent Zn–Cd interaction outcomes under controlled co-exposure conditions, rather than to define a general agronomic Zn supplementation threshold or a full Zn dose–response relationship. On the final day before sampling, photosynthetic parameters and chlorophyll fluorescence were measured. Samples were harvested on day 15 after treatment. Plant height, biomass, and active root surface area were measured immediately after harvest, after which each sample was divided into three portions for the determination of metal concentrations, physiological indices, and molecular analyses, respectively. Experimental design and sample stratification The core phenotyping experiment included 44 rice cultivars under two treatments (Cd alone and Cd + Zn), with three biological replicates per cultivar per treatment. At harvest, all cultivars were sampled using the same biological replicate framework. Photosynthetic traits, chlorophyll fluorescence, growth parameters, root activity, and elemental concentrations were measured across the full 44-cultivar panel. In parallel, backup frozen samples were collected for all cultivars at the same sampling point. To enable deeper mechanistic profiling while controlling analytical cost, 12 representative cultivars were selected from the 44-cultivar panel for subsequent targeted assays. These cultivars were chosen to cover all four response quadrants identified in the Zn–Cd response landscape and to represent contrasting response types across the panel. The 12-cultivar subset was used for the mechanistically enriched measurements included in the sPLS analysis, including ROS/antioxidant traits, hormone and flavonoid/phenolic profiles, photosynthetic pigment variables, and transporter-expression-related data. Photosynthetic and Chlorophyll Fluorescence Measurements Photosynthetic parameters and chlorophyll fluorescence were measured using a portable photosynthesis system (LI-6400XT, LI-COR, USA) and a chlorophyll fluorometer (JUNIOR-PAM, WALZ, Germany), respectively. Photosynthetic measurements were performed under artificial light (~ 25,000 lux, equivalent to ~ 462.5 µmol photons m⁻² s⁻¹) using ambient-air open-flow conditions and the instrument's standard operating settings. All samples were measured within the same time window under identical measurement settings to minimize environmental variation among cultivars. Active Root Surface Area Analysis After 14 days of treatment, fresh rice roots were subjected to TTC staining to evaluate root activity according to the method described by Wang et al.(2023) with slight modifications. Briefly, 0.5 g of fresh roots was incubated in 5 mL of 0.4% (w/v) 2,3,5-triphenyltetrazolium chloride (TTC) solution at 37°C in the dark for 2 hours. The reaction was terminated by adding 2 mL of 1 M H₂SO₄, and roots were rinsed thoroughly with deionized water. To quantify active root surface area, stained roots were scanned at 400 dpi using a root analysis system (WinRHIZO, Regent Instruments, Canada). Active root surface area was determined by applying a standardized color threshold corresponding to TTC staining. Metal Concentration Analysis The determination of Cd and Zn concentrations was performed according to the method of Ilieva et al. (Ilieva, Angelova, Drochioiu, Murariu & Surleva 2019), with slight modifications. Dried samples were ground to a fine powder using a stainless-steel grinder and passed through a 100-mesh (150 µm) nylon sieve. Approximately 0.3 g (± 0.1 mg) of the powdered sample was weighed into quartz digestion vessels and treated with 8 mL of concentrated nitric acid (65%). Digestion was carried out using a super microwave digestion system (SUPEC EXPEC 790S, Expec-Tech, China) with the following program: ramping from 25°C to 240°C over 20 minutes, holding at 240°C for 30 minutes, then cooling to 50°C. Digested samples were filtered through 0.45 µm PTFE membranes, diluted to 50 mL with ultrapure water, and analyzed for Cd and Zn concentrations using flame atomic absorption spectrophotometry (AAS; ICE-3500, Thermo Fisher Scientific, USA). Endogenous Hormones and Flavonoids Analysis Quantification of endogenous hormones and flavonoids followed an adapted LC-MS/MS method referencing protocols by Park et al. (Park et al., 2017 ) and Mustafa et al. (Mustafa et al., 2022 ). Analytical standards were sourced from Sigma-Aldrich and Merck (USA), with solvents and reagents, including formic acid, methanol, and acetonitrile, obtained from Fisher Scientific (USA). For extraction, approximately 100 mg of frozen sample was homogenized in 80% methanol, then centrifuged and filtered through a 0.22 µm membrane before LC-MS/MS analysis. A Vanquish UHPLC coupled with a Quantis TSQ mass spectrometer (Thermo Fisher Scientific, USA) facilitated chromatographic separation using a C18 column under gradient conditions (10–90% acetonitrile containing 0.1% formic acid; flow rate of 0.3 mL/min; temperature at 30°C). Detection was performed using electrospray ionization (ESI) and multiple reaction monitoring (MRM) with instrument settings optimized as detailed in Table S1 . Measurement of Reactive Oxygen Species and Antioxidant Enzyme Activities Hydrogen peroxide (H₂O₂) and superoxide anion (O₂•⁻) levels, as well as the activities of peroxidase (POD), superoxide dismutase (SOD), and catalase (CAT), were determined using commercially available assay kits according to the manufacturers' protocols, with minor modifications to sample extraction and handling. For ROS measurements, approximately 0.1 g of tissue was homogenized in liquid nitrogen, extracted overnight in 80% methanol at 4°C, and centrifuged at 10,000 × g for 10 min to obtain the supernatant for analysis. H₂O₂ and O₂•⁻ contents were then quantified spectrophotometrically using the corresponding kits (Solarbio, China). For antioxidant enzyme assays, approximately 0.5 g of tissue was homogenized in chilled phosphate buffer (pH 7.0) and centrifuged at 12,000 × g for 20 min to obtain crude enzyme extracts. POD, SOD, and CAT activities were measured using commercial assay kits (Sangon Biotech, China) based on guaiacol oxidation, inhibition of nitroblue tetrazolium photoreduction, and H₂O₂ decomposition, respectively. Absorbance was recorded at 470 nm for POD, 560 nm for SOD, and 240 nm for CAT using a spectrophotometer (Presee T9CS, China). Protein concentration was determined by the Bradford method, and enzyme activities were normalized to protein. Gene Expression Analysis by RT-qPCR Gene expression levels of metal transporter genes in rice were quantified by real-time reverse transcription quantitative PCR (RT-qPCR). Twelve representative cultivars were selected from the 44-cultivar panel based on the Zn–Cd response landscape to cover all four response quadrants and contrasting response types. Root samples from these 12 cultivars were used for transporter-expression analysis to provide targeted support for the multivariate candidate signal structures inferred from the sPLS analysis. Total RNA was extracted from root tissues using a Polysaccharide and Polyphenol Plant RNA Extraction Kit (Servicebio, China). RNA concentration and purity were assessed using a NanoDrop spectrophotometer (Thermo Scientific, USA). First-strand cDNA was synthesized from 1 µg of total RNA using the MightyScript First Strand cDNA Synthesis Master Mix (Sangon Biotech, China) according to the manufacturer's instructions and then diluted tenfold with nuclease-free water. RT-qPCR was performed using the 2× SG Fast qPCR Master Mix (Sangon Biotech, China) on a CFX Opus 96 Real-Time PCR System (Bio-Rad, USA). The rice actin gene ( OsACT1 ) was used as the reference gene (Liu et al., 2023 ; Wang et al., 2016 ). Gene-specific primers targeting key Cd- and Zn-related transporter genes are listed in Table S2. Primer specificity was verified by melting curve analysis (65–95°C), and amplification efficiency was determined using standard curves generated from fivefold serial dilutions of cDNA. Relative transcript levels were calculated using the 2 −ΔΔCt method. Each sample included three biological replicates and three technical replicates. Data analysis All statistical analyses and visualizations were performed in R (version 4.5.3). Pearson correlation coefficients were calculated to assess associations between baseline metal concentrations and Zn-induced physiological response variables. To quantify the directional and magnitude diversity of Zn–Cd interactions across cultivars, Log₂ fold changes (Log₂FC) of root and shoot Cd concentrations between Zn + Cd and Cd-only treatments were calculated for all 44 cultivars and used to construct a two-dimensional response landscape. Response intensity was defined as the Euclidean distance of each cultivar's coordinate from the origin, and cultivars were classified into four response quadrants based on the sign of root and shoot Log₂FC. Random Forest models were constructed using the ranger package (v0.18.0) with 1,000 trees and permutation-based variable importance. Two classification models (M1 and M2) were fitted to predict the direction of Zn-induced changes in shoot Cd accumulation (LCd_delta class) and root-to-shoot translocation efficiency (TF_delta class), respectively. Model performance was evaluated using out-of-bag (OOB) predictions, and accuracy and balanced accuracy were calculated from OOB confusion matrices. To assess whether classification performance exceeded random expectation, permutation tests were performed for M1 and M2 by repeatedly shuffling class labels (1,000 permutations) and recalculating model performance metrics; empirical one-sided p-values were obtained by comparing the observed accuracy, balanced accuracy, AUC, and kappa values against the corresponding null distributions. Two regression models (M3 and M4) were fitted to predict the magnitude of Zn-induced changes in total Cd accumulation (ΔTotalCd) and translocation efficiency (ΔTF), respectively, with observed-versus-predicted R², RMSE, and MAE used to summarize model performance. For all models, predictor variables comprised baseline metal-status traits (leaf and root Cd and Zn concentrations under Cd-only conditions) and Zn-induced physiological response variables (Z − C differences in growth and photochemical parameters). The mtry parameter was set to floor( \(\:\sqrt{p}\) ) for classification models and floor(p/3) for regression models, where p denotes the number of predictor variables. Two additional sensitivity models (M1-alt and M2-alt) were fitted using absolute endpoint values as response variables to assess whether the prominence of baseline leaf Cd concentration reflected arithmetic coupling rather than an independent biological signal. Following the Random Forest analysis, two-way ANOVA was additionally applied using base R functions to a targeted subset of core metal-status, growth, and photochemical traits to assess whether these variables also showed significant cultivar, treatment, and cultivar × treatment effects from a complementary univariate, design-based perspective. P values were adjusted using the Benjamini–Hochberg method within each effect category. The RF layer was used to identify informative cultivar-level predictors associated with response direction or magnitude at the population scale. In contrast, the downstream sPLS layer was applied to the representative 12-cultivar subset to resolve deeper multivariable structures underlying specific response dimensions. Sparse partial least squares (sPLS) regression was performed using the mixOmics package (v6.34.0) on a subset of 12 representative cultivars selected from the 44-cultivar panel to cover all four Zn–Cd response quadrants and contrasting response types. Three models were constructed to represent distinct response layers: S1 for Zn-induced changes in translocation efficiency (TF_delta), S2 for changes in leaf Cd accumulation (LCd_delta), and S3 for changes in total Cd accumulation (totalCd_delta). The predictor matrix comprised Z − C delta values of a mechanistically enriched variable set encompassing ROS/antioxidant traits, HMA transporter expression, phenolics/flavonoids, photosynthetic pigments, and hormones. All predictor variables were centered and scaled before model fitting. Optimal numbers of components and retained variables per component were selected via leave-one-out cross-validation using the tune.spls function with Pearson correlation as the optimization criterion; final models were fitted with two components and five retained variables per component. For structural interpretation, retained-variable identities and loading magnitudes were extracted from each component of the three final sPLS models and compared across models. Variables retained uniquely in one response layer, or repeatedly contributing to one model but not the others, were used to describe layer-enriched candidate structures and to assess whether TF_delta-, LCd_delta-, and totalCd_delta-associated signal spaces showed substantial overlap or relative partitioning at the retained-variable level. Given the limited sample size and the absence of usable Q² output under the current cross-validation setting, sPLS models were interpreted as candidate-signal extraction tools rather than generalized predictive frameworks, and the cross-model comparison was treated descriptively rather than as formal evidence of mechanistic independence. Results Baseline Metal Loading Is Associated with the Directional and Physiological Response of Rice to Zn–Cd Interaction: A Quantitative Landscape Analysis To quantify the directional and magnitude diversity of Zn–Cd interactions across rice genotypes, Log₂ fold changes (Log₂FC) of Cd concentrations in roots and shoots between Cd-only and Zn + Cd treatments were calculated for 44 cultivars and visualized as a response landscape (Fig. 1 a). Overall response intensity, defined as the Euclidean distance of each cultivar's coordinate from the origin, ranged from 0.02 to 3.89 (mean = 1.05), indicating substantial genotypic variation in response to the tested Zn co-exposure condition. Based on the sign of root and shoot Log₂FC, cultivars were classified into four response modes: Dual Inhibition (n = 14), in which both root and shoot Cd declined simultaneously; Shoot Promotion/Root Inhibition (n = 13), characterized by decreased root Cd alongside increased shoot Cd; Root Promotion/Shoot Inhibition (n = 11), showing the opposite pattern; and Dual Promotion (n = 6), in which Cd increased in both organs. Cultivar 7 exhibited the highest overall intensity (3.89), driven primarily by strong shoot suppression (shoot Log₂FC = − 3.88), whereas the root response was comparatively modest (root Log₂FC = − 0.28). Cultivar 43 (intensity = 2.87) occupied the upper-left extreme of the Shoot Promotion quadrant, while cultivar 27 was the most distal representative of the Dual Promotion quadrant. Beyond Cd accumulation, under the tested Zn + Cd co-exposure condition, Zn addition was accompanied by widespread yet highly divergent physiological responses across all 44 cultivars (Fig. 1 b). Among photosystem indicators, Y(II) increased in 30 out of 44 cultivars (68.2%; median Δ = +0.016), and ETR showed a similar trend (65.9% positive; median Δ = +3.05), suggesting a general tendency toward improved photochemical efficiency, albeit with considerable inter-cultivar variation. In contrast, qP decreased in the majority of cultivars (27/44, 61.4%; median Δ = −0.037), indicating that photochemical quenching capacity responded more heterogeneously across genotypes. For growth-related traits, root fresh weight declined in 31 of 44 cultivars (70.5%; median Δ = −0.25 g), representing the most consistent directional response across the dataset, while shoot fresh weight and total biomass also showed predominantly negative shifts (61.4% and 63.6%, respectively). Shoot water content was a notable exception, increasing in 63.6% of cultivars (median Δ = +2.80%). Root activity also showed substantial genotype-dependent variability (SD = 7.24). Although 61.4% of cultivars exhibited a positive response, with a median Δ of + 2.37, the mean remained slightly negative (− 0.36), indicating that a subset of cultivars experienced pronounced declines that offset the more common moderate increases. Collectively, these patterns indicate that Zn addition triggers highly genotype-specific physiological shifts across photosystem efficiency, biomass allocation, root activity, and water status, thereby defining a multidimensional response background against which Zn–Cd interaction types can be further resolved by random forest analysis. (a) Log₂ fold change (Log₂FC) of root and shoot Cd concentrations between Zn + Cd and Cd-only treatments, plotted as a two-dimensional response landscape for 44 rice cultivars. Each point represents one cultivar; point size and color indicate response intensity (Euclidean distance from the origin: Low, Medium, High). Dashed lines demarcate four response quadrants: Dual Promotion (root↑ shoot↑, n = 6), Shoot Promotion/Root Inhibition (root↓ shoot↑, n = 13), Dual Inhibition (root↓ shoot↓, n = 14), and Root Promotion/Shoot Inhibition (root↑ shoot↓, n = 11). Selected cultivar IDs are labeled. (b) Strip plots showing the distribution of Δ values (Zn + Cd minus Cd-only) for 11 physiological variables across all 44 cultivars. Each dot represents one cultivar; red = positive response (increase), blue = negative response (decrease). The solid vertical line in each panel indicates the median; the dashed line marks zero (no change). Panels are grouped by module: photosystem indicators (Fv/Fm, Y(II), ETR, qP, Y(NPQ); blue background), root physiology (Root Activity; green background), and growth traits (Biomass, Root FW, Shoot FW, Root WC, Shoot WC; orange background). Random Forest analysis identifies cultivar-specific Zn–Cd response types and their associated predictors. To dissect the factors underlying the divergent Zn–Cd interaction patterns across cultivars, we constructed four primary Random Forest (RF) models that address both response direction and magnitude. M1 and M2 were classification models used to predict whether Zn induced a promoted or inhibited response in shoot Cd accumulation and root-to-shoot translocation, respectively, with M2 serving as the primary model for translocation-related response types. M3 and M4 were regression models used to quantify the magnitude of Zn-mediated changes in total Cd accumulation (Δ𝑇𝑜𝑡𝑎𝑙𝐶𝑑) and translocation efficiency (Δ𝑇𝐹). In all models, predictors were structured into two layers: baseline metal-status traits, representing the pre-existing elemental chassis, and physiological response traits, representing Zn-induced functional shifts. To assess whether the arithmetic relationship between the baseline trait L Cd Conc_C and the delta-based response variables inflated their variable importance rankings, two sensitivity models (M1-alt and M2-alt) were additionally fitted using absolute endpoint values (L Cd Conc_Z and TF_Z, respectively) as response variables, with identical predictor sets. This design provided a common framework for evaluating the joint contribution of elemental background and physiological response to cultivar-specific Zn–Cd interaction phenotypes. Model performance evaluation revealed a clear contrast between the predictability of response direction and that of response magnitude. The two classification models showed moderate but meaningful discriminative capacity, with M1 (shoot Cd response) and M2 (translocation response) reaching accuracies of 70.45% and 68.18%, and balanced accuracies of 68.32% and 67.91%, respectively. By contrast, the two regression models explained only 13.79% (M3) and 23.82% (M4) of the variance. This overall pattern indicates that the Zn-induced Cd response was more readily classified in terms of direction than quantitatively predicted in terms of magnitude. Permutation tests further indicated that the observed classification performance was unlikely to be entirely due to random label structure. For M1, both accuracy (0.6818 vs. null mean 0.4954, permutation p = 0.032) and balanced accuracy (0.6632 vs. null mean 0.4744, permutation p = 0.030) exceeded random expectation. For M2, balanced accuracy also marginally exceeded the null distribution (0.6553 vs. null mean 0.4781, permutation p = 0.050). AUC-based comparisons, however, were weaker for both models and did not clearly exceed the corresponding null distributions. Together, these results support the interpretation that the RF models captured modest but non-random population-level structure in response-direction classification, particularly for shoot Cd response direction. Variable-importance analysis revealed distinct but complementary patterns across the four models. In M1 (Fig. 2 a), L Cd Conc_C ranked first (0.0359), followed by Biomass (Z-C) and L Zn Conc_C, with Y(II) (Z-C) and Fv/Fm (Z-C) also retained among the top five. M2 (Fig. 2 c) showed a closely parallel structure, with L Cd Conc_C again ranking first (0.0423) and L Zn Conc_C, Y(II) (Z-C), Biomass (Z-C), and No. plants (Z-C) in the following order. The convergence of predictor structures between M1 and M2 suggests that the factors associated with shoot Cd accumulation direction and translocation response direction were largely shared across cultivars, consistent with translocation being a primary determinant of shoot Cd accumulation under the present experimental conditions. Of note, L Cd Conc_C ranked first in both models. However, because it contributes arithmetically to both delta-based response variables, its prominence may partly reflect structural coupling rather than fully independent biological information. Sensitivity models showed that removing this coupling did not disrupt the broader ranking structure, but redistributed importance toward different subsets of predictors (Fig. S1 ). Although the sensitivity models showed lower overall predictive performance, consistent with the greater background variance inherent in absolute endpoint values (M1-alt OOB R² = 0.061; M2-alt OOB R² = 0.106; Fig. S1 a-b), the consistency of key predictors across primary and sensitivity models supports the robustness of the observed importance patterns. In M1-alt (Fig. 2 b), Biomass (Z-C) and Fv/Fm (Z-C) emerged as the leading predictors, suggesting that cultivar-specific physiological responses to Zn were informative for classifying whether shoot Cd accumulation was promoted or inhibited. In M2-alt (Fig. 2 d), L Zn Conc_C retained high importance alongside several root metal-status traits, and the shift from a predominantly leaf-based predictor structure in M2 to a more root-enriched structure in M2-alt suggests that the direction of translocation response and the absolute efficiency of root-to-shoot Cd transfer were associated with partially distinct organ-level predictor structures. In the regression models, L Zn Conc_C and R Zn Conc_C were the strongest predictors of translocation magnitude in M4 (Fig. S2c), suggesting that the extent of Zn-induced changes in translocation efficiency was closely associated with the pre-existing Zn accumulation status of individual cultivars. In M3 (Fig. S2a), physiological response variables contributed relatively little, suggesting that prediction of Zn-induced changes in total Cd uptake relied more strongly on baseline metal-status traits than on the physiological response variables included here. Together, these patterns indicate that baseline elemental status and Zn-induced physiological shifts jointly contributed to cultivar-specific Zn–Cd interaction phenotypes, but with stronger relevance for response classification than for quantitative prediction of response magnitude. The regression models showed weaker overall explanatory performance. The low predictive power of both M3 and M4 indicates that the absolute magnitude of the Zn–Cd interaction could not be robustly captured by the current variable set alone, and several cultivars showed large residuals in M4, indicating substantial model-unexplained heterogeneity among genotypes. Together, these results suggest that the variables used here were more effective in discriminating response type than in quantitatively resolving response intensity. Horizontal bars represent permutation importance values for each predictor variable. (a) M1: classification model for shoot Cd response direction (response = LCd_delta class; n = 44). (b) M1-alt: sensitivity model using absolute shoot Cd concentration as the response variable (response = L Cd Conc_Z; n = 44). (c) M2: classification model for translocation response direction (response = TF_delta class; n = 44). (d) M2-alt: sensitivity model using absolute translocation factor as the response variable (response = TF_Z; n = 44). All four models used identical predictor sets comprising baseline metal-status traits (L/R Cd and Zn concentrations under control conditions) and Zn-induced physiological response traits (Z − C differences in growth and photochemical parameters). Negative importance values indicate that including the variable degraded model predictions. M1-alt and M2-alt were fitted to assess whether the prominence of L Cd Conc_C in M1 and M2 reflected arithmetic coupling with the delta-based response variables rather than an independent biological signal. Complementary univariate analyses further supported the RF framework. Across the selected follow-up traits, two-way ANOVA consistently detected significant cultivar effects and cultivar × treatment interactions after BH correction (Table S3), indicating that Zn-induced variation in metal status, growth, and photochemical traits was strongly genotype-dependent. Treatment main effects were also significant for most variables except leaf Cd concentration. sPLS analysis reveals layer-specific multivariable structures underlying Zn-induced Cd response variation To further resolve the deep-layer signals associated with cultivar-specific Zn–Cd interaction patterns, we performed sparse partial least squares (sPLS) analyses using the 12 selected cultivars and a mechanistically enriched variable set, including ROS/antioxidant traits, HMA transporter expression, phenolics/flavonoids, pigments, and hormones. Three models were constructed to represent different response layers: S1 for Zn-induced changes in translocation efficiency (TF_delta), S2 for changes in leaf Cd accumulation (LCd_delta), and S3 for changes in total Cd accumulation (totalCd_delta). All models were fitted with two components and five retained variables per component, yielding similarly high fitting performance (R² = 0.904, 0.789, and 0.872 for S1, S2, and S3, respectively; observed-versus-predicted plots shown in Figure S3). However, given the limited sample size and absence of Q² values, these models were interpreted primarily as tools for candidate-signal extraction rather than generalized prediction. Accordingly, the 12-cultivar sPLS analysis was positioned as a deeper interpretive layer that complemented rather than replaced the 44-cultivar Random Forest framework. Among the three models, S1 showed the most integrative structure for Cd translocation response (Fig. 3 a, b). Its retained variables spanned two distinct dimensions. Comp1 was dominated by R_O 2 • − _delta (loading = -0.778), with R_Astragalin_delta (loading = -0.606) providing additional phenolic/flavonoid-related structure, and R_HMA2_delta, L_Astragalin_delta, and R_POD_delta contributing smaller supplementary signals. This primary axis indicated that Zn-induced root superoxide and flavonoid responses were the major features associated with separation along the primary translocation-related latent dimension. Comp2 revealed a secondary dimension dominated by R_SA_delta (loading = -0.831), with R_ HMA1 _delta (loading = 0.467) and L_O 2 • − _delta (loading = 0.256) contributing additional transporter- and antioxidant-associated structure, indicating that salicylate-related reconfiguration defined a secondary hormone-associated axis within the translocation layer. S2 revealed a different pattern for Zn-induced changes in leaf Cd accumulation (Fig. 3 c, d). Comp1 was driven primarily by flavonoid- and hormone-related variables, with L_Quercetin_delta showing the strongest loading (0.734) and R_JA_delta contributing substantial additional support (loading = 0.652); L_SL_delta, L_Cinnamic_Acid_delta, and R_ HMA1 _delta were retained with smaller positive loadings. Comp2 defined a distinct secondary axis dominated by R_Astragalin_delta (loading = -0.752), with L_Protamine_Sulfate_delta (loading = 0.480) and Carotenoid_delta (loading = -0.439) as the next strongest contributors, and L_SOD_delta and L_H 2 O 2 _delta providing minor antioxidant-associated signals. Together, the two components indicated that leaf quercetin/jasmonate responses defined the primary leaf Cd accumulation axis, whereas a secondary dimension captured root astragalin and leaf carotenoid co-variation. S3, in turn, showed a third-variable structure in total Cd response amplitude (Fig. 3 e, f). Comp1 was overwhelmingly dominated by R_Quinic_Acid_delta (loading = 0.928), with R_Quercetin_delta retained at a moderate negative loading (-0.363) and L_Rosmarinic_Acid_delta, L_Quercetin_delta, and R_CAT_delta contributing minor supplementary signals. This primary axis indicated that root quinic acid accumulation was by far the strongest feature associated with total Cd response amplitude. Comp2 revealed a qualitatively distinct secondary dimension centered on HMA transporter expression, with R_ HMA7 _delta as the dominant variable (loading = 0.890) and R_HMA3_delta showing a moderate negative loading (-0.403); R_JA_delta and R_CAT_delta contributed smaller hormone- and antioxidant-associated signals. This pattern suggests that while phenolic reconfiguration defined the primary axis of total Cd response variation, a secondary HMA7 / HMA3 -associated transporter dimension was also captured within the same model. Viewed together, the three sPLS models revealed related but non-identical multivariable structures across response layers. At the comp1 level, the dominant retained variables showed clear layer-specific partitioning: S1 was centered on a root superoxide/flavonoid-associated structure (R_O 2 • − _delta, R_Astragalin_delta), S2 on a quercetin/jasmonate-associated structure (L_Quercetin_delta, R_JA_delta), and S3 on a root quinic acid-dominated phenolic structure (R_Quinic_Acid_delta). At the comp2 level, each model captured a qualitatively distinct secondary dimension: a salicylate/ HMA -associated axis in S1, an astragalin/carotenoid-associated axis in S2, and an HMA7 / HMA3 transporter axis in S3. Notably, R_O 2 • − _delta was dominant in S1 comp1 but absent from S2 and S3 comp1, while HMA7-associated signals appeared as a secondary dimension in S3 but not as a primary feature in either S1 or S2. This limited overlap across both components indicates that Zn-induced variation in translocation efficiency, leaf Cd accumulation, and total Cd accumulation was associated with partially distinct candidate structures rather than a single shared latent mechanism. (a) Sample scores of S1, the sPLS model for Zn-induced changes in translocation efficiency (TF_delta). (b) Retained variables in S1. (c) Sample scores of S2, the sPLS model for Zn-induced changes in leaf Cd accumulation (LCd_delta). (d) Retained variables in S2. (e) Sample scores of S3, the sPLS model for Zn-induced changes in total Cd accumulation (totalCd_delta). (f) Retained variables in S3. Across the three models, the retained-variable structures were partially overlapping but non-identical. S1 comp1 was dominated by root superoxide and flavonoid signals (R_O 2 • − _delta, R_Astragalin_delta), with a secondary salicylate/HMA-associated axis (R_SA_delta, R_ HMA1 _delta) in comp2. S2 comp1 was defined by leaf quercetin and jasmonate responses (L_Quercetin_delta, R_JA_delta), with a secondary astragalin/carotenoid axis (R_Astragalin_delta, Carotenoid_delta) in comp2. S3 comp1 was overwhelmingly dominated by root quinic acid (R_Quinic_Acid_delta), with a secondary HMA transporter-associated axis (R_HMA7_delta, R_ HMA3 _delta) in comp2. Discussion The present study indicates that Zn-Cd interaction in rice cannot be described adequately as a uniformly antagonistic process under the tested near-equimolar co-exposure regime (Cai et al., 2019 ; Wang et al., 2024 ). Across 44 cultivars, the response landscape revealed a four-quadrant distribution of interaction types and a nearly 200-fold range in response intensity, showing that Zn-associated changes in Cd accumulation varied substantially among genotypes (Tavarez et al., 2023 , 2022 ). Beyond Cd redistribution, Zn addition was accompanied by divergent shifts in photosystem efficiency, biomass allocation, root activity, and water status, indicating that the observed heterogeneity spans multiple physiological dimensions rather than a single trait (Chang et al., 2023 ). This heterogeneity was not random with respect to the measured predictor set: Random Forest models classified response direction with approximately 70% accuracy using baseline metal status and Zn-induced physiological shifts (Wright and Ziegler, 2017 ). These results do not establish causality, but they do indicate that response direction is associated with partially discriminable biological features captured by the present framework. Taken together, the data support treating Zn-Cd interaction outcome as a structured, genotype-associated response phenotype and justify analyzing response direction and response intensity as related but partially distinct dimensions rather than collapsing them into a single response metric (Mashabela et al., 2023 ). Beyond discriminating response direction, the RF models revealed several coarse-grained interpretive patterns with independent analytical value. First, the near-identical predictor structures of M1 and M2—with L Cd Conc_C ranking first in both models and Zn-induced physiological shifts (Biomass Z-C, Y(II) Z-C, Fv/Fm Z-C) consistently retained among the top predictors—indicate that shoot Cd accumulation direction and translocation response direction are associated with a largely shared set of cultivar-level features, suggesting that the genotypic features associated with whether Zn promotes or inhibits root-to-shoot translocation substantially overlap with those associated with shoot Cd accumulation direction—consistent with translocation being the dominant process through which genotypic identity may shape shoot Cd outcomes, with little evidence of systematic decoupling between translocation and leaf accumulation responses across the cultivar panel (Uraguchi et al., 2009 ). Second, variable importance analysis in M4 revealed that baseline Zn status in both leaves and roots (L Zn Conc_C, importance = 0.659; R Zn Conc_C, importance = 0.536) was the dominant predictor of translocation magnitude, substantially outranking all physiological response variables, several of which contributed negatively (Fv/Fm Z-C: −0.054; Biomass Z-C: −0.136). This pattern indicates that the extent of Zn-induced translocation shifts is more closely tied to a cultivar's pre-existing Zn accumulation status than to its observable physiological response trajectory. We therefore infer that baseline Zn homeostatic status may be associated with translocation response amplitude, possibly reflecting differences in cellular redox and metal-sensing states relevant to downstream transporter responses—a candidate hypothesis examined in the sPLS analysis below. Third, the shift from a predominantly leaf-based predictor structure in M2 to a more root-enriched structure in M2-alt suggests that the directional and quantitative dimensions of translocation response are associated with partially distinct organ-level informational layers—with root metal status more closely linked to absolute translocation capacity (Uraguchi et al., 2009 ) and leaf physiological variables more informative for classifying response direction, possibly reflecting systemic differentiation between shoot and root response layers. Whether this organ-level asymmetry reflects shoot-to-root signaling, shared genotypic architecture expressed differently across organs, or both, remains to be determined, and is examined at the candidate-signal level in the sPLS analysis below. Permutation-based null comparisons further supported the interpretation that the RF models captured non-random discriminative structure rather than purely chance separation. However, this signal was stronger for shoot Cd response direction than for translocation response direction, and remained modest in probability-ranking terms as reflected by the weaker AUC-based comparisons. A key interpretive implication of the three-tier framework concerns the relationship between the RF-detectable surface signal and the sPLS-resolved candidate structure. The RF layer captured cultivar-level predictors that were informative for response classification, but did not by itself determine whether photochemical, transporter-related, or redox-associated variables should be regarded as primary or secondary within a given response layer (Mashabela et al., 2023 ). The sPLS comparison helped reduce this ambiguity by showing that the apparently informative physiological signal did not behave as a single shared mechanistic axis across all response dimensions (Chun and Keleş, 2010 ). Instead, translocation-related variation was associated mainly with a root superoxide/flavonoid structure, with R_O 2 • − _delta and R_Astragalin_delta forming the dominant comp1 variables and additional comp2 contributions from salicylate- and transporter-related variables (Kostyuk et al., 2004 ), whereas leaf Cd accumulation was associated more closely with a quercetin/jasmonate co-varying structure, with L_Quercetin_delta and R_JA_delta dominating comp1 and additional comp2 contributions from astragalin- and carotenoid-related variables (Lei et al., 2020 ). This pattern is consistent with the interpretation that flavonoid and hormonal signals function as layer-dependent phenotypic readouts of deeper metal-homeostasis/redox states rather than as a single shared upstream driver (Considine and Foyer, 2014 ). Taken together, these RF-derived patterns highlight three coarse-grained population-level inferences: translocation appears to be the primary axis associated with shoot Cd response direction; baseline Zn homeostasis is associated with translocation response magnitude; and root- and leaf-derived predictor layers were differentially associated with translocation direction versus efficiency. However, the regression models also revealed a clear resolution limit: M4 explained only 23.8% of the variance in translocation magnitude, and M3 explained only 13.8% of the variance in total Cd uptake. The genotypes with the most extreme translocation responses were precisely those least captured by the present variable set, suggesting that intensity outliers may harbor biological features qualitatively absent from the surface-level predictor space. The limited predictive power of the regression models indicates that RF analysis across 44 cultivars functioned primarily as a coarse-grained filter, sufficient to capture the directional patterning of Zn–Cd interaction phenotypes but unable to resolve the finer-grained biological variation underlying response intensity. These follow-up univariate results reinforce the interpretation that Zn–Cd response diversity is genotype-dependent not only at the multivariate pattern-recognition level captured by RF, but also at the level of conventional trait-by-treatment interaction. The variables associated with response intensity—including HMA transporter expression, ROS/antioxidant status, hormonal signaling, and secondary metabolite profiles—operate at a deeper biological resolution than can be captured by population-scale physiological screening alone. The sPLS analysis on the 12 representative cultivars was therefore designed as a fine-grained filtering step operating on a qualitatively distinct variable set, thereby extending the analysis into a deeper candidate-signal space encompassing HMA transporter expression, ROS/antioxidant status, hormonal signaling, and secondary metabolite profiles, and identifying layer-specific candidate signal structures as testable hypotheses. The sPLS analysis conducted on the 12 representative cultivars indicated that Zn–Cd interaction diversity is multivariate and response-layer specific: no single variable dominated across all three models, and the retained signal structures of S1 (translocation efficiency, TF_delta), S2 (leaf Cd accumulation, LCd_delta), and S3 (total Cd accumulation, totalCd_delta) showed limited overlap at the dominant-variable level. Rather than converging on a single shared latent structure, the three response layers were associated with distinct candidate axes, supporting the view that Zn-induced changes in translocation, leaf accumulation, and total accumulation are organized by partially separable biological structures (Mashabela et al., 2023 ). At the translocation layer (S1), the dominant retained variables point to a root redox/flavonoid-associated structure rather than to a chlorophyll- or pigment-dominated one. Comp1 was dominated by R_O 2 • − _delta (loading = -0.778) and R_Astragalin_delta (loading = -0.606), indicating that Zn-induced root superoxide reconfiguration and astragalin-related flavonoid shifts were the most prominent features associated with the primary translocation axis in the present dataset (Feigl et al., 2015 ; Tian et al., 2025 ). A secondary dimension represented by comp2 additionally retained strong salicylate-related variation, with R_SA_delta showing the largest loading, while HMA-related transporter signals, including R_ HMA2 _delta, contributed further transporter-associated structure(Kim and Jang, 2009 ; Lee et al., 2007 ; Sasaki et al., 2014 ; Ueno et al., 2010 ). Because these relationships derive from retained-variable composition and loading structure rather than direct perturbation, they are best interpreted as candidate features of a translocation-associated redox/flavonoid axis rather than as proof of a causal sequence. Integrating the S1 observations, we propose a candidate redox-baseline framework in which cultivar-specific root redox state under Cd-alone conditions may influence the amplitude of Zn-induced redox reorganization and thereby contribute to variation in Zn-Cd interaction direction (Cai et al., 2019 ; Umair Hassan et al., 2020 ). In this view, redox and flavonoid status would act not as isolated explanatory variables, but as parts of a broader translocation-associated candidate structure that also includes salicylate- and HMA-related transporter responses. This interpretation remains provisional and should be tested in broader germplasm panels and targeted perturbation experiments. At the leaf Cd accumulation layer (S2), the predictor structure shifted toward a quercetin/jasmonate-associated organization. Within the current retained-variable structure, comp1 was dominated by L_Quercetin_delta (loading = 0.734) and R_JA_delta (loading = 0.652), with L_SL_delta and L_Cinnamic_Acid_delta providing additional phenolic/hormonal support, indicating that Zn-induced leaf flavonoid accumulation and root jasmonate signaling were the primary features associated with the leaf Cd accumulation axis (Chen et al., 2021 ). A secondary dimension, represented by comp2, additionally retained R_Astragalin_delta and Carotenoid_delta, along with minor antioxidant-associated signals, suggesting that a partially distinct astragalin/carotenoid co-varying dimension also contributed to the leaf accumulation layer (Ramel et al., 2012 ). Relative to S1, this pattern indicates that leaf Cd accumulation was associated less with the root superoxide-dominated structure characterizing the translocation layer, and more with a flavonoid/hormone-centered axis in which quercetin and jasmonate responses co-varied with secondary phenolic and pigment shifts (B. Wang et al., 2023 ). The present data support association, not directional causation, and therefore do not establish whether the quercetin- and jasmonate-related signals are upstream regulators of Cd partitioning, downstream stress readouts, or parallel components of a broader leaf-level response structure. The difference between S1 and S2 is important because it suggests that the physiological signals retained within the broader analytical framework should not be interpreted as a single universal mechanism across all Zn-Cd response dimensions (Mashabela et al., 2023 ). Instead, their interpretive meaning appears to depend on which response layer is being examined. In the present dataset, the strongest translocation-associated signal was linked to a root superoxide/flavonoid structure centered on R_O 2 • − _delta and R_Astragalin_delta (Zhang et al., 2023 ). In contrast, leaf Cd accumulation aligned more closely with a quercetin/jasmonate co-varying structure, where flavonoid and hormonal shifts likely reflect leaf-level stress responses accompanying Cd partitioning rather than serving as direct Cd transport routes (Lei et al., 2020 ). This layer-specific partitioning strengthens the view that RF-level physiological signals are better interpreted as entry points into deeper, layer-specific candidate structures than as self-sufficient mechanistic explanations. The present dataset also suggests broader implications for the evaluation of Zn-based mitigation strategies in rice. Because Zn was used here as a model co-exposure factor rather than a field-optimized agronomic amendment, the present results do not justify general claims about the universal safety or efficacy of Zn supplementation (Tavarez et al., 2023 ; S. Wang et al., 2023 ). They do, however, indicate that cultivar background may condition whether Zn co-exposure is associated with inhibition, promotion, or more complex redistribution of Cd under a fixed near-equimolar regime (Cai et al., 2019 ; Tavarez et al., 2023 ). This raises the possibility that genotype-dependent responsiveness may represent a broader candidate dimension in rice Cd-risk evaluation. However, that broader applicability remains to be tested before any screening or management application is claimed. An additional implication of this framework is that constitutive low-Cd tendency and co-exposure responsiveness may not fully overlap as cultivar dimensions (Sasaki et al., 2014 ; Tavarez et al., 2023 ). Traits that reduce constitutive Cd entry or retention (Sasaki et al., 2014 ; Ueno et al., 2010 ) would be expected to lower oxidative burden under Cd stress, whereas the present data suggest that a stronger root redox baseline state may be associated with a more responsive translocation-related layer under Zn-Cd co-exposure (Uraguchi et al., 2009 ; Xue et al., 2023 ). Although this possibility is not tested directly here, it raises the question of whether cultivars optimized for low constitutive Cd accumulation and those showing stronger responsiveness to external modulation necessarily represent the same physiological strategy. The layer-specific signal structures identified across S1, S2, and S3 also raise the question of whether the organ-level informational asymmetry detected in the RF models - where leaf baseline traits dominated some coarse-grained classifications, yet root redox or phenolic variables dominated several sPLS latent axes - reflects a general feature of genotype-dependent Zn–Cd interaction or only the specific multivariate composition of the present cultivar subset (Tavarez et al., 2023 ; Uraguchi et al., 2009 ). At a minimum, the current results show that different response layers need not converge on the same dominant retained variables. That apparent cross-organ inconsistency at the RF layer can be reconciled by a deeper candidate-structure view in which different organs contribute differently to different response dimensions(Chun and Keleş, 2010 ; Mashabela et al., 2023 ). Notwithstanding these layer-specific patterns, a subset of cultivars — including cultivars 27, 43, 10, and 11 — repeatedly deviated from the dominant latent structures across one or more sPLS models, as reflected in their sample score distributions and observed-versus-predicted residuals. This residual heterogeneity indicates that the candidate signal profiles identified here, while capturing the dominant axes of variation among the 12 cultivars, do not fully account for the biological complexity present in a subset of genotypes. Cultivars 27 and 43, in particular, showed hormone-rich profiles with coordinated HMA and antioxidant responses that partially overlapped across multiple signal layers, suggesting that these genotypes may integrate multiple regulatory modules more synergistically than the layer-specific sPLS structures can resolve individually. It should be noted that the sPLS models were interpreted as candidate signal-extraction tools rather than generalized predictive frameworks, given the limited sample size of 12 cultivars and the absence of usable cross-validation Q² values. The candidate signal patterns identified here, therefore, require validation in larger, more diverse germplasm panels before their generalizability can be established. Conclusion In summary, under a fixed near-equimolar Zn + Cd co-exposure condition, rice cultivars exhibited substantial diversity in both the direction and magnitude of Zn-associated changes in Cd accumulation. By integrating response-landscape analysis, Random Forest classification, and sPLS-based candidate-structure extraction, this study shows that response direction was more tractable than response magnitude, and that distinct Zn–Cd response layers were associated with partially differentiated multivariable structures. Translocation-related variation was linked primarily to a root superoxide/flavonoid-associated structure, leaf Cd accumulation to a more flavonoid/hormone-centered candidate structure, and total Cd amplitude to a phenolic/flavonoid-associated structure. Together, these results support a layer-specific candidate framework for genotype-dependent Zn–Cd interaction, rather than a single shared mechanism, and highlight root redox baseline state under Cd stress alone as a potential upstream feature associated with response direction. More broadly, the study suggests that constitutive Cd accumulation tendency and co-exposure responsiveness may represent only partially overlapping cultivar dimensions. Declarations Conflict of Interest On behalf of all authors, the corresponding author states that there is no conflict of interest. Ethics, Consent to Participate, and Consent to Publish declarations Not applicable. Funding This research was supported by the Natural Science Foundation of National Natural Science Foundation of China (32201392), and Guangxi Major Science and Technology Program (Guike AA24263045). Author Contribution S.C. conceived the study, curated the data, developed the methodology, and drafted the manuscript. B.S. contributed to methodology development and software analysis. F.Z. contributed to the investigation and resources. Y.L. contributed to data curation and visualization. X.C. contributed to visualization. Q.L. contributed to data curation, funding acquisition, project administration, and manuscript review and editing. All authors have read and agreed to the published version of the manuscript. Acknowledgements The authors would like to thank all who contributed to this work. Data Availability All data generated or analyzed during this study will be available upon reasonable request. References Adil MF, Sehar S, Chen G, Chen Z-H, Jilani G, Chaudhry AN, Shamsi IH (2020) Cadmium-zinc cross-talk delineates toxicity tolerance in rice via differential genes expression and physiological / ultrastructural adjustments. Ecotoxicol Environ Saf 190:110076. https://doi.org/10.1016/j.ecoenv.2019.110076 Cai Y, Xu W, Wang M, Chen W, Li X, Li Y, Cai Y (2019) Mechanisms and uncertainties of Zn supply on regulating rice Cd uptake. Environ Pollut Barking Essex 1987 253:959–965. https://doi.org/10.1016/j.envpol.2019.07.077 Chang W, Wang W, Shi Z, Cao G, Zhao X, Su X, Chen Y, Wu J, Yang Z, Liu C, Shang L, Cai Z (2023) Comparative metabolomics combined with physiological analysis revealed cadmium tolerance mechanism in indica rice ( Oryza sativa L). J Agric Food Chem 71:7669–7678. https://doi.org/10.1021/acs.jafc.3c00850 Chen X, Jiang W, Tong T, Chen G, Zeng F, Jang S, Gao W, Li Z, Mak M, Deng F, Chen Z-H (2021) Molecular interaction and evolution of jasmonate signaling with transport and detoxification of heavy metals and metalloids in plants. Front Plant Sci 12:665842. https://doi.org/10.3389/fpls.2021.665842 Chun H, Keleş S (2010) Sparse partial least squares regression for simultaneous dimension reduction and variable selection. J R Stat Soc Ser B Stat Methodol 72:3–25. https://doi.org/10.1111/j.1467-9868.2009.00723.x Considine MJ, Foyer CH (2014) Redox regulation of plant development. Antioxid Redox Signal 21:1305–1326. https://doi.org/10.1089/ars.2013.5665 Du J, Zeng J, Ming X, He Q, Tao Q, Jiang M, Gao S, Li X, Lei T, Pan Y, Chen Q, Liu S, Yu X (2020) The presence of zinc reduced cadmium uptake and translocation in Cosmos bipinnatus seedlings under cadmium/zinc combined stress. Plant Physiol Biochem 151:223–232. https://doi.org/10.1016/j.plaphy.2020.03.019 Feigl G, Lehotai N, Molnár Á, Ördög A, Rodríguez-Ruiz M, Palma JM, Corpas FJ, Erdei L, Kolbert Z (2015) Zinc induces distinct changes in the metabolism of reactive oxygen and nitrogen species (ROS and RNS) in the roots of two brassica species with different sensitivity to zinc stress. Ann Bot 116:613–625. https://doi.org/10.1093/aob/mcu246 Fontanili L, Lancilli C, Suzui N, Dendena B, Yin Y-G, Ferri A, Ishii S, Kawachi N, Lucchini G, Fujimaki S, Sacchi GA, Nocito FF (2016) Kinetic analysis of zinc/cadmium reciprocal competitions suggests a possible Zn-insensitive pathway for root-to-shoot cadmium translocation in rice. Rice 9:16–28. https://doi.org/10.1186/s12284-016-0088-3 Haider FU, Liqun C, Coulter JA, Cheema SA, Wu J, Zhang R, Wenjun M, Farooq M (2021) Cadmium toxicity in plants: impacts and remediation strategies. Ecotoxicol Environ Saf 211:111887. https://doi.org/10.1016/j.ecoenv.2020.111887 Hu J (2021) Toward unzipping the ZIP metal transporters: structure, evolution, and implications on drug discovery against cancer. FEBS J 288:5805–5825. https://doi.org/10.1111/febs.15658 Hussain B, Ashraf MN, Shafeeq-Ur-Rahman N, Abbas A, Li J, Farooq M (2021) Cadmium stress in paddy fields: effects of soil conditions and remediation strategies. Sci Total Environ 754:142188. https://doi.org/10.1016/j.scitotenv.2020.142188 Kim G-N, Jang H-D (2009) Protective mechanism of quercetin and rutin using glutathione metabolism on HO-induced oxidative stress in HepG2 cells. Ann N Y Acad Sci 1171:530–537. https://doi.org/10.1111/j.1749-6632.2009.04690.x Kostyuk VA, Potapovich AI, Strigunova EN, Kostyuk TV, Afanas'ev IB (2004) Experimental evidence that flavonoid metal complexes may act as mimics of superoxide dismutase. Arch Biochem Biophys 428:204–208. https://doi.org/10.1016/j.abb.2004.06.008 Lee S, Kim Y-Y, Lee Y, An G (2007) Rice P1B-type heavy-metal ATPase, OsHMA9, is a metal efflux protein. Plant Physiol 145:831–842. https://doi.org/10.1104/pp.107.102236 Lei GJ, Sun L, Sun Y, Zhu XF, Li GX, Zheng SJ (2020) Jasmonic acid alleviates cadmium toxicity in arabidopsis via suppression of cadmium uptake and translocation. J Integr Plant Biol 62:218–227. https://doi.org/10.1111/jipb.12801 Liu X, Gao Y, Zhao X, Zhang X, Ben L, Li Z, Dong G, Zhou J, Huang J, Yao Y (2023) Validation of novel reference genes in different rice plant tissues through mining RNA-seq datasets. Plants 12:3946. https://doi.org/10.3390/plants12233946 Mashabela MD, Masamba P, Kappo AP (2023) Applications of metabolomics for the elucidation of abiotic stress tolerance in plants: a special focus on osmotic stress and heavy metal toxicity. Plants 12:269–286. https://doi.org/10.3390/plants12020269 Mustafa AM, Abouelenein D, Angeloni S, Maggi F, Navarini L, Sagratini G, Santanatoglia A, Torregiani E, Vittori S, Caprioli G (2022) A New HPLC-MS/MS Method for the Simultaneous Determination of Quercetin and Its Derivatives in Green Coffee Beans. Foods 11:3033. https://doi.org/10.3390/foods11193033 Park CH, Yeo HJ, Park YJ, Morgan AMA, Valan Arasu M, Al-Dhabi NA, Park SU (2017) Influence of Indole-3-Acetic Acid and Gibberellic Acid on Phenylpropanoid Accumulation in Common Buckwheat (Fagopyrum esculentum Moench) Sprouts. Molecules 22:374. https://doi.org/10.3390/molecules22030374 Qixing Z, Yanyu W, Xianzhe X (1994) Compound pollution of Cd and Zn and its ecological effect on rice plant. Chin J Appl Ecol 5:438–441 Ramel F, Birtic S, Ginies C, Soubigou-Taconnat L, Triantaphylidès C, Havaux M (2012) Carotenoid oxidation products are stress signals that mediate gene responses to singlet oxygen in plants. Proc. Natl. Acad. Sci. 109, 5535–5540. https://doi.org/10.1073/pnas.1115982109 Rizwan M, Ali S, Rehman MZ, ur, Maqbool A (2019) A critical review on the effects of zinc at toxic levels of cadmium in plants. Environ Sci Pollut Res 26:6279–6289. https://doi.org/10.1007/s11356-019-04174-6 Sasaki A, Yamaji N, Ma JF (2014) Overexpression of OsHMA3 enhances Cd tolerance and expression of Zn transporter genes in rice. J Exp Bot 65:6013–6021. https://doi.org/10.1093/jxb/eru340 Shahzad M, Bibi A, Khan A, Shahzad A, Xu Z, Maruza TM, Zhang G (2025) Utilization of antagonistic interactions between micronutrients and cadmium (Cd) to alleviate Cd toxicity and accumulation in crops. Plants 14:707–721. https://doi.org/10.3390/plants14050707 Tavarez M, Grusak MA, Sankaran RP (2023) The effect of exogenous cadmium and zinc applications on cadmium, zinc and essential mineral bioaccessibility in three lines of rice that differ in grain cadmium accumulation. Foods 12:4026. https://doi.org/10.3390/foods12214026 Tavarez M, Grusak MA, Sankaran RP (2022) Effects of zinc fertilization on grain cadmium accumulation, gene expression, and essential mineral partitioning in rice. Agronomy 12:2182. https://doi.org/10.3390/agronomy12092182 Tian X, Zhang J, Ye Z, Fang W, Ding X, Yin Y (2025) Zinc sulfate stress enhances flavonoid content and antioxidant capacity from finger millet sprouts for high-quality production. Foods 14:2563. https://doi.org/10.3390/foods14152563 Ueno D, Yamaji N, Kono I, Huang CF, Ando T, Yano M, Ma JF (2010) Gene limiting cadmium accumulation in rice. Proc. Natl. Acad. Sci. 107, 16500–16505. https://doi.org/10.1073/pnas.1005396107 Umair Hassan M, Aamer M, Umer Chattha M, Haiying T, Shahzad B, Barbanti L, Nawaz M, Rasheed A, Afzal A, Liu Y, Guoqin H (2020) The critical role of zinc in plants facing the drought stress. Agriculture 10:396–415. https://doi.org/10.3390/agriculture10090396 Uraguchi S, Mori S, Kuramata M, Kawasaki A, Arao T, Ishikawa S (2009) Root-to-shoot Cd translocation via the xylem is the major process determining shoot and grain cadmium accumulation in rice. J Exp Bot 60:2677–2688. https://doi.org/10.1093/jxb/erp119 Vance TM, Chun OK (2015) Zinc intake is associated with lower cadmium burden in US adults. J Nutr 145:2741–2748. https://doi.org/10.3945/jn.115.223099 Vu KT, Lan PDT, Nguyen NTH, Thanh HN (2022) Cadmium immobilization in the rice - paddy soil with biochar additive 23. 85–89. https://doi.org/10.12911/22998993/146331 Wang B, Lin L, Yuan X, Zhu Y, Wang Y, Li D, He J, Xiao Y (2023) Low-level cadmium exposure induced hormesis in peppermint young plant by constantly activating antioxidant activity based on physiological and transcriptomic analyses. Front Plant Sci 14. https://doi.org/10.3389/fpls.2023.1088285 Wang H, Liu M, Zhang Y, Jiang Q, Wang Q, Gu Y, Song X, Li Y, Ye Y, Wang F, Chen X, Wang Z (2024) Foliar spraying of Zn/Si affects Cd accumulation in paddy grains by regulating the remobilization and transport of Cd in vegetative organs. Plant Physiol Biochem 207:108351. https://doi.org/10.1016/j.plaphy.2024.108351 Wang H, Xu C, Luo Z-C, Zhu H-H, Wang S, Zhu Q-H, Huang D-Y, Zhang Y-Z, Xiong J, He Y-B (2018) Foliar application of Zn can reduce Cd concentrations in rice ( Oryza sativa L.) under field conditions. Environ Sci Pollut Res Int 25:29287–29294. https://doi.org/10.1007/s11356-018-2938-6 Wang Q, Zeng X, Song Q, Sun Y, Feng Y, Lai Y (2020) Identification of key genes and modules in response to cadmium stress in different rice varieties and stem nodes by weighted gene co-expression network analysis. Sci Rep 10:9525. https://doi.org/10.1038/s41598-020-66132-4 Wang S, Wu M, Zhong S, Sun J, Mao X, Qiu N, Zhou F (2023) A rapid and quantitative method for determining seed viability using 2,3,5-triphenyl tetrazolium chloride (TTC): with the example of wheat seed. Molecules 28:6828. https://doi.org/10.3390/molecules28196828 Wang Z, Wang Y, Yang J, Hu K, An B, Deng X, Li Y (2016) Reliable selection and holistic stability evaluation of reference genes for rice under 22 different experimental conditions. Appl Biochem Biotechnol 179:753–775. https://doi.org/10.1007/s12010-016-2029-4 Wen B, Jiang H, Gao Y, Zhou Q, Qie H (2024) Source analysis and bioavailability of soil cadmium in poyang lake plain of China based on principal component analysis and positive definite matrix factor. Minerals 14:514–526. https://doi.org/10.3390/min14050514 Wright MN, Ziegler A (2017) ranger: a fast implementation of random forests for high dimensional data in C + + and. R J Stat Softw 77:1–17. https://doi.org/10.18637/jss.v077.i01 Xue W, Zhang X, Zhang C, Wang C, Huang Y, Liu Z (2023) Mitigating the toxicity of reactive oxygen species induced by cadmium via restoring citrate valve and improving the stability of enzyme structure in rice. Chemosphere 327:138511. https://doi.org/10.1016/j.chemosphere.2023.138511 Yoshida S, Forno D, Cock J (1971) Laboratory manual for physiological studies of rice, Laboratory manual for physiological studies of rice. Los Baños, Philippines Zhang H, Sun X, Hwarari D, Du X, Wang Y, Xu H, Lv S, Wang T, Yang L, Hou D (2023) Oxidative stress response and metal transport in roots of macleaya cordata exposed to lead and zinc. Plants 12:516. https://doi.org/10.3390/plants12030516 Zhang S, Wu X, Peng J, Meng X, Shi B, Zhou L, Bai L (2021) Study of the physiological dynamics of cadmium accumulation in two varieties of rice with different cadmium-accumulating properties. J. Chem. 2021, 6238893. https://doi.org/10.1155/2021/6238893 Additional Declarations No competing interests reported. Supplementary Files Supplement.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9361934","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":621974402,"identity":"9bc9b450-2a41-4f0d-9d6d-d2cecfd7e19d","order_by":0,"name":"Shimiao Chen","email":"","orcid":"","institution":"Guangxi Subtropical Crops Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Shimiao","middleName":"","lastName":"Chen","suffix":""},{"id":621974403,"identity":"fc5af488-fcc0-473a-a147-1ca2d50a090f","order_by":1,"name":"Bin Shan","email":"","orcid":"","institution":"Ministry of Agriculture and Rural Affairs","correspondingAuthor":false,"prefix":"","firstName":"Bin","middleName":"","lastName":"Shan","suffix":""},{"id":621974404,"identity":"31c3e6a5-ee55-401b-a295-087f9af2ab30","order_by":2,"name":"Fuhai Zheng","email":"","orcid":"","institution":"Guangxi Academy of Agricultural Science","correspondingAuthor":false,"prefix":"","firstName":"Fuhai","middleName":"","lastName":"Zheng","suffix":""},{"id":621974405,"identity":"e4917cdf-d02c-4b4f-9ffe-e6b002abace9","order_by":3,"name":"Yanyan Li","email":"","orcid":"","institution":"Qinzhou Institute of Agricultural Sciences","correspondingAuthor":false,"prefix":"","firstName":"Yanyan","middleName":"","lastName":"Li","suffix":""},{"id":621974406,"identity":"58a03520-c1cc-4cff-965c-fa5ec8d7f5b7","order_by":4,"name":"Xi Chen","email":"","orcid":"","institution":"Henan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Xi","middleName":"","lastName":"Chen","suffix":""},{"id":621974407,"identity":"c1de5498-2127-4efb-a730-de5e663e2571","order_by":5,"name":"Qinyu Lu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxklEQVRIiWNgGAWjYBACPiA+AMRyxGthg2oxhovwEKMFBBIbiNcikbzxcGHb4fT5EckPP/5guCNnT1hLWsHhmW2HczeeOWYszcPwzJgIW3IMDvOCtLQ3GEgzMBxO7CFWS7phM/vnnz8YDtcTrSVBnr3HTIKH4XACYYfxPCs4zHMu3XADz5kyax6Dw4Y9Bwho4WdP3vyZp8xaXn5G+uabPyoOy7M3ELKGgcGAgREYOwYHIGyiAFDZHwYGeSIMHwWjYBSMghEKAGk5PRqbd9lQAAAAAElFTkSuQmCC","orcid":"","institution":"Guangxi Academy of Agricultural Science","correspondingAuthor":true,"prefix":"","firstName":"Qinyu","middleName":"","lastName":"Lu","suffix":""}],"badges":[],"createdAt":"2026-04-09 01:53:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9361934/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9361934/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107186094,"identity":"8bccd302-2c39-495e-9455-cf686962aa91","added_by":"auto","created_at":"2026-04-17 18:57:04","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":87228,"visible":true,"origin":"","legend":"\u003cp\u003eQuantitative Zn–Cd response landscape and physiological response overview of 44 rice cultivars.\u003c/p\u003e\n\u003cp\u003e(a) Log₂ fold change (Log₂FC) of root and shoot Cd concentrations between Zn+Cd and Cd-only treatments, plotted as a two-dimensional response landscape for 44 rice cultivars. Each point represents one cultivar; point size and color indicate response intensity (Euclidean distance from the origin: Low, Medium, High). Dashed lines demarcate four response quadrants: Dual Promotion (root↑ shoot↑, n = 6), Shoot Promotion/Root Inhibition (root↓ shoot↑, n = 13), Dual Inhibition (root↓ shoot↓, n = 14), and Root Promotion/Shoot Inhibition (root↑ shoot↓, n = 11). Selected cultivar IDs are labeled. (b) Strip plots showing the distribution of Δ values (Zn+Cd minus Cd-only) for 11 physiological variables across all 44 cultivars. Each dot represents one cultivar; red = positive response (increase), blue = negative response (decrease). The solid vertical line in each panel indicates the median; the dashed line marks zero (no change). Panels are grouped by module: photosystem indicators (Fv/Fm, Y(II), ETR, qP, Y(NPQ); blue background), root physiology (Root Activity; green background), and growth traits (Biomass, Root FW, Shoot FW, Root WC, Shoot WC; orange background).\u003c/p\u003e","description":"","filename":"Picture1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9361934/v1/4e11ad3c6b506ee81835c447.jpg"},{"id":107483300,"identity":"7c19793d-abfc-4b49-94f3-67973c1e471a","added_by":"auto","created_at":"2026-04-22 02:27:16","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":38485,"visible":true,"origin":"","legend":"\u003cp\u003eVariable importance rankings from primary and sensitivity Random Forest models for shoot Cd response (M1/M1-alt) and root-to-shoot translocation response (M2/M2-alt).\u003c/p\u003e\n\u003cp\u003eHorizontal bars represent permutation importance values for each predictor variable. (a) M1: classification model for shoot Cd response direction (response = LCd_delta class; n = 44). (b) M1-alt: sensitivity model using absolute shoot Cd concentration as the response variable (response = L Cd Conc_Z; n = 44). (c) M2: classification model for translocation response direction (response = TF_delta class; n = 44). (d) M2-alt: sensitivity model using absolute translocation factor as the response variable (response = TF_Z; n = 44). All four models used identical predictor sets comprising baseline metal-status traits (L/R Cd and Zn concentrations under control conditions) and Zn-induced physiological response traits (Z − C differences in growth and photochemical parameters). Negative importance values indicate that including the variable degraded model predictions. M1-alt and M2-alt were fitted to assess whether the prominence of L Cd Conc_C in M1 and M2 reflected arithmetic coupling with the delta-based response variables rather than an independent biological signal. Complementary univariate analyses further supported the RF framework. Across the selected follow-up traits, two-way ANOVA consistently detected significant cultivar effects and cultivar × treatment interactions after BH correction (Table S3), indicating that Zn-induced variation in metal status, growth, and photochemical traits was strongly genotype-dependent. Treatment main effects were also significant for most variables except leaf Cd concentration.\u003c/p\u003e","description":"","filename":"Picture2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9361934/v1/c0bcff3ca90bdd4f4608c8ba.jpg"},{"id":107186096,"identity":"8c6f45e7-e7eb-4937-94e0-6db5c002a425","added_by":"auto","created_at":"2026-04-17 18:57:04","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":43109,"visible":true,"origin":"","legend":"\u003cp\u003eLayer-specific candidate structures identified by sPLS across three Cd-response layers.\u003c/p\u003e\n\u003cp\u003e(a) Sample scores of S1, the sPLS model for Zn-induced changes in translocation efficiency (TF_delta). (b) Retained variables in S1. (c) Sample scores of S2, the sPLS model for Zn-induced changes in leaf Cd accumulation (LCd_delta). (d) Retained variables in S2. (e) Sample scores of S3, the sPLS model for Zn-induced changes in total Cd accumulation (totalCd_delta). (f) Retained variables in S3. Across the three models, the retained-variable structures were partially overlapping but non-identical. S1 comp1 was dominated by root superoxide and flavonoid signals (R_O\u003csub\u003e2\u003c/sub\u003e•\u003csup\u003e-\u003c/sup\u003e_delta, R_Astragalin_delta), with a secondary salicylate/HMA-associated axis (R_SA_delta, R_\u003cem\u003eHMA1\u003c/em\u003e_delta) in comp2. S2 comp1 was defined by leaf quercetin and jasmonate responses (L_Quercetin_delta, R_JA_delta), with a secondary astragalin/carotenoid axis (R_Astragalin_delta, Carotenoid_delta) in comp2. S3 comp1 was overwhelmingly dominated by root quinic acid (R_Quinic_Acid_delta), with a secondary HMA transporter-associated axis (R_HMA7_delta, R_\u003cem\u003eHMA3\u003c/em\u003e_delta) in comp2.\u003c/p\u003e","description":"","filename":"Picture3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9361934/v1/592a69ff8dcb3e40935893d8.jpg"},{"id":108264935,"identity":"3a9806b5-bfcb-4c4e-b8ce-543e06156613","added_by":"auto","created_at":"2026-05-01 09:25:37","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":516679,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9361934/v1/205081e7-ff84-4130-aa65-3c1ebab22ff6.pdf"},{"id":107186093,"identity":"d2504dc0-4aa3-443b-a083-4e405e631d02","added_by":"auto","created_at":"2026-04-17 18:57:04","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":1233981,"visible":true,"origin":"","legend":"","description":"","filename":"Supplement.docx","url":"https://assets-eu.researchsquare.com/files/rs-9361934/v1/101323821a2c2054a5493544.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Genotype-Dependent Dual Effects of Zinc on Cadmium Accumulation in Rice Revealed by a Multi-Scale Quantitative Framework","fulltext":[{"header":"Introduction","content":"\u003cp\u003eZinc (Zn) and cadmium (Cd) co-occur widely in agricultural soils, particularly in paddy systems affected by phosphate fertilizer inputs, mining activities, or industrial emissions (Hussain et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Because Zn\u0026sup2;⁺ and Cd\u0026sup2;⁺ share similar ionic radii and charge properties, they compete for common uptake and translocation pathways, including ZIP-family transporters and low-affinity cation channels such as calcium-permeable channels (Cai et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Hu, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). This structural and chemical similarity means that elevated Zn availability can saturate shared transport systems, effectively reducing Cd influx into root cells\u0026mdash;the mechanistic basis for antagonistic Zn\u0026ndash;Cd interactions (Wang et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Under such conditions, Zn supplementation, including foliar application of ZnSO₄, has been shown to reduce Cd concentrations in plant tissues and grain substantially, and to enhance antioxidant enzyme activities that mitigate Cd-induced oxidative stress (Rizwan et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Beyond competitive inhibition at the transporter level, Zn may also influence rhizosphere chemistry, altering soil pH and organic matter interactions and thereby modifying Cd speciation and bioavailability before root uptake (Du et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). However, the outcome of Zn\u0026ndash;Cd co-exposure is not uniformly antagonistic, and its direction is strongly dependent on the relative concentrations and molar ratios of the two metals, as well as on soil chemical properties and plant physiological status (Cai et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Under Zn-deficient conditions, reduced competitive inhibition can paradoxically enhance Cd uptake and translocation, yielding synergistic interactions and elevated Cd burden in plant tissues (Haider et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Even under adequate Zn supply, certain cultivar-specific responses defy the simple antagonism model: Zhou et al. observed that simultaneous Cd\u0026ndash;Zn contamination suppressed Zn uptake while paradoxically elevating Cd accumulation in rice tissues, illustrating that the directionality of the interaction cannot be predicted from metal concentrations alone (Shahzad et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Zn supplementation may further modify the bioaccessibility and gastrointestinal solubility of grain Cd \u0026mdash; through alterations in Cd speciation and binding within grain tissues \u0026mdash; thereby altering actual human exposure risk in ways that total Cd concentration alone would fail to capture (Vance and Chun, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Collectively, these bidirectional, ratio-dependent, and context-sensitive outcomes reflect an interaction system whose direction and magnitude are jointly governed by soil chemistry, metal speciation, plant physiology, and cultivation context (Du et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Qixing et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e1994\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe complexity of Zn\u0026ndash;Cd interactions is further shaped by the substantial genotypic variation that exists among rice cultivars in their capacity to take up, translocate, and sequester Cd. Rice (\u003cem\u003eOryza sativa\u003c/em\u003e L.) is the primary dietary Cd exposure route for over half the global population (Vu et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and its efficient root-to-grain Cd translocation makes it a crop of particular concern under contaminated field conditions (Zhang et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Cd-tolerant genotypes typically exhibit more robust antioxidant defense systems, more efficient Cd immobilization in the cell wall fraction, and greater vacuolar sequestration capacity, collectively reducing Cd translocation to aboveground tissues and limiting its deposition in grain (Chang et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Comparative metabolomic profiling of contrasting indica varieties \u0026mdash; the Cd-tolerant NH224 and the Cd-sensitive NH199 \u0026mdash; demonstrated that tolerance was associated with enhanced amino acid biosynthesis, elevated hormone metabolism, strengthened phenylpropanoid pathway activity, and upregulated antioxidant enzyme systems, illustrating the breadth of metabolic reprogramming that underlies genotypic Cd resilience (Chang et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). At the molecular level, this variation is mechanistically anchored in the differential expression and functional activity of a small set of membrane transporters. OsNramp5, localized to root epidermal cells, mediates initial Cd entry from the rhizosphere and represents the primary influx route; \u003cem\u003eOsHMA2\u003c/em\u003e, expressed in root vascular tissue, controls xylem loading and root-to-shoot translocation; and \u003cem\u003eOsHMA3\u003c/em\u003e, a tonoplast-localized vacuolar transporter highly expressed in root cells, sequesters Cd into vacuoles and thereby restricts its upward movement toward aerial tissues and grain (Ueno et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Allelic polymorphisms and genotype-specific transcriptional regulation of these transporters \u0026mdash; modulated by Cd-responsive signaling pathways, phytohormone fluctuations, and oxidative stress cues \u0026mdash; drive marked cultivar-level differences in Cd partitioning (Sasaki et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Cultivars harboring functional OsHMA3 alleles exhibit substantially lower shoot and grain Cd accumulation compared to those lacking them, underscoring the central role of vacuolar sequestration capacity as a determinant of grain safety (Sasaki et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Critically, because these same transporter systems mediate both baseline Cd handling and the competitive dynamics between Zn\u0026sup2;⁺ and Cd\u0026sup2;⁺ at shared binding sites, genotypic background is likely a primary determinant of how a given cultivar responds to Zn co-exposure \u0026mdash; yet this connection has rarely been examined systematically (Adil et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Tavarez et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite this mechanistic foundation, understanding of how Zn\u0026ndash;Cd interactions manifest across cultivated rice germplasm remains limited and fragmented. Existing studies have reported both antagonistic and synergistic outcomes. Still, these observations are usually confined to individual cultivars or narrow experimental settings, and the reported directions of interaction often contradict one another across studies (Tavarez et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Some studies describe robust suppression of grain Cd following Zn supplementation. In contrast, others report paradoxical increases in grain Cd in particular genotypes, with these discrepancies attributed to differences in root exudation, rhizosphere pH, transporter affinity, or cultivar-specific metal partitioning (Fontanili et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Tavarez et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Interpretation is further complicated by substantial methodological heterogeneity, including differences in Zn and Cd concentrations, Zn/Cd molar ratios, treatment timing, and the tissues selected for metal quantification, all of which make direct cross-study comparison difficult (Shahzad et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Crucially, no study to date has systematically characterized the full spectrum of Zn\u0026ndash;Cd interaction outcomes \u0026mdash; from strong antagonism to strong synergism \u0026mdash; across a representative cultivar panel under controlled and directly comparable exposure conditions (Shahzad et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Tavarez et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). As a result, it remains unclear whether the contradictory findings in the literature primarily reflect genuine biological diversity among genotypes, methodological heterogeneity, or both. Equally lacking is a structured framework for identifying the cultivar-level biological features associated with whether Zn supplementation inhibits or promotes Cd accumulation, and for determining whether response direction and response magnitude are governed by the same or distinct biological axes.\u003c/p\u003e \u003cp\u003eTwo specific and interrelated questions, therefore, remain unresolved. First, which cultivar-level biological features - encompassing baseline metal accumulation status, Zn-induced shifts in transporter expression, redox and antioxidant responses, hormonal signaling, and secondary-metabolite mobilization - are associated with whether a genotype responds to Zn co-exposure with net Cd inhibition or net Cd enhancement (Shahzad et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e)? If response direction is associated with measurable cultivar traits, it becomes possible to stratify germplasm by likely interaction pattern under defined co-exposure conditions and to assess whether cultivar background should be considered when interpreting Zn-based mitigation outcomes. More broadly, this also raises the question of how reliably Zn-based mitigation effects can be generalized across cultivar backgrounds, because conclusions drawn from single-cultivar or narrow-panel studies may not fully capture the diversity of outcomes observed in broader rice germplasm (Shahzad et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Tavarez et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Second, are the biological features associated with response direction the same as those that govern response magnitude, or do these represent partially distinct biological axes? This distinction matters because direction and amplitude may reflect different physiological processes operating at different regulatory levels. Disentangling these two dimensions requires an analytical framework capable of handling multivariate, multi-layered biological signals simultaneously rather than examining single variables or pathways in isolation (Mashabela et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Resolving these questions is important both for building a mechanistic understanding of cultivar-specific Zn-Cd interaction and for defining testable candidate features that may condition interaction outcome under combined exposure (Hussain et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Zhang et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo address these gaps, the present study investigated Zn-Cd interaction patterns across 44 rice cultivars under a fixed near-equimolar co-exposure regime. This design was chosen not to characterize the full agronomic dose-dependence of Zn supplementation, but to enable a controlled, directly comparable assessment of cultivar-specific interaction outcomes under equivalent molar inputs of the two metals. An integrated three-tier analytical framework was employed. A quantitative response landscape was first constructed to characterize the directional and magnitude diversity of Zn-Cd interactions at the population level across all 44 cultivars. Random Forest modeling was then applied to identify the cultivar-level features - comprising baseline metal status and Zn-induced physiological shifts - that were most informative for response direction and magnitude. Finally, sparse partial least squares analysis was performed on a subset of 12 representative cultivars using a mechanistically enriched variable set to identify candidate signal structures associated with distinct Zn-Cd response layers, including transporter expression, ROS/antioxidant status, hormonal signaling, and secondary metabolite profiles (Chun and Keleş, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). This multi-scale approach was designed to characterize population-level diversity in Zn\u0026ndash;Cd interaction outcomes and to define candidate biological features associated with response direction and magnitude under the tested co-exposure condition, while also providing a broader framework for understanding cultivar-specific differences in constitutive Cd accumulation tendency and co-exposure responsiveness.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePlant materials, treatments, and sampling\u003c/h2\u003e \u003cp\u003eForty-four rice varieties (\u003cem\u003eOryza sativa\u003c/em\u003e L.), comprising Indica and Japonica genotypes, were sourced from local germplasm collections and commercially cultivated varieties widely planted across Guangxi, China (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Seeds were surface-sterilized with 2.5% sodium hypochlorite for 10 min and rinsed thoroughly with distilled water before germination. Seeds were germinated in the dark at 28\u0026deg;C for 2 days, then sown onto the seedling substrate. Uniform seedlings were transferred to a half-strength Yoshida nutrient solution (1971) for hydroponic cultivation at the one-leaf-one-heart stage. After reaching the three-leaf\u0026ndash;one-heart stage (~\u0026thinsp;7 days), Cd and Zn treatments were applied. The nutrient solution was aerated continuously, refreshed every 3 days, and maintained at pH 5.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1. The samples were then divided into two groups. The control group was grown in full-strength Yoshida nutrient solution containing 0.2 mg/L Cd alone. In comparison, the treatment group was simultaneously exposed to 0.2 mg/L Cd and 0.12 mg/L Zn for 14 days under controlled hydroponic conditions. The Cd concentration (0.2 mg/L) was selected based on previously reported contamination-relevant exposure levels(Wen et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The Zn concentration (0.12 mg/L) was chosen to provide a near-equimolar co-exposure relative to Cd, so that the two metals were supplied at comparable molar intensity within a fixed dual-metal regime. This design was intended to facilitate interpretation of competitive and genotype-dependent Zn\u0026ndash;Cd interaction outcomes under controlled co-exposure conditions, rather than to define a general agronomic Zn supplementation threshold or a full Zn dose\u0026ndash;response relationship.\u003c/p\u003e \u003cp\u003eOn the final day before sampling, photosynthetic parameters and chlorophyll fluorescence were measured. Samples were harvested on day 15 after treatment. Plant height, biomass, and active root surface area were measured immediately after harvest, after which each sample was divided into three portions for the determination of metal concentrations, physiological indices, and molecular analyses, respectively.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eExperimental design and sample stratification\u003c/h3\u003e\n\u003cp\u003eThe core phenotyping experiment included 44 rice cultivars under two treatments (Cd alone and Cd\u0026thinsp;+\u0026thinsp;Zn), with three biological replicates per cultivar per treatment. At harvest, all cultivars were sampled using the same biological replicate framework. Photosynthetic traits, chlorophyll fluorescence, growth parameters, root activity, and elemental concentrations were measured across the full 44-cultivar panel. In parallel, backup frozen samples were collected for all cultivars at the same sampling point. To enable deeper mechanistic profiling while controlling analytical cost, 12 representative cultivars were selected from the 44-cultivar panel for subsequent targeted assays. These cultivars were chosen to cover all four response quadrants identified in the Zn\u0026ndash;Cd response landscape and to represent contrasting response types across the panel. The 12-cultivar subset was used for the mechanistically enriched measurements included in the sPLS analysis, including ROS/antioxidant traits, hormone and flavonoid/phenolic profiles, photosynthetic pigment variables, and transporter-expression-related data.\u003c/p\u003e\n\u003ch3\u003ePhotosynthetic and Chlorophyll Fluorescence Measurements\u003c/h3\u003e\n\u003cp\u003ePhotosynthetic parameters and chlorophyll fluorescence were measured using a portable photosynthesis system (LI-6400XT, LI-COR, USA) and a chlorophyll fluorometer (JUNIOR-PAM, WALZ, Germany), respectively. Photosynthetic measurements were performed under artificial light (~\u0026thinsp;25,000 lux, equivalent to ~\u0026thinsp;462.5 \u0026micro;mol photons m⁻\u0026sup2; s⁻\u0026sup1;) using ambient-air open-flow conditions and the instrument's standard operating settings. All samples were measured within the same time window under identical measurement settings to minimize environmental variation among cultivars.\u003c/p\u003e\n\u003ch3\u003eActive Root Surface Area Analysis\u003c/h3\u003e\n\u003cp\u003e After 14 days of treatment, fresh rice roots were subjected to TTC staining to evaluate root activity according to the method described by Wang et al.(2023) with slight modifications. Briefly, 0.5 g of fresh roots was incubated in 5 mL of 0.4% (w/v) 2,3,5-triphenyltetrazolium chloride (TTC) solution at 37\u0026deg;C in the dark for 2 hours. The reaction was terminated by adding 2 mL of 1 M H₂SO₄, and roots were rinsed thoroughly with deionized water. To quantify active root surface area, stained roots were scanned at 400 dpi using a root analysis system (WinRHIZO, Regent Instruments, Canada). Active root surface area was determined by applying a standardized color threshold corresponding to TTC staining.\u003c/p\u003e\n\u003ch3\u003eMetal Concentration Analysis\u003c/h3\u003e\n\u003cp\u003eThe determination of Cd and Zn concentrations was performed according to the method of Ilieva et al. (Ilieva, Angelova, Drochioiu, Murariu \u0026amp; Surleva 2019), with slight modifications. Dried samples were ground to a fine powder using a stainless-steel grinder and passed through a 100-mesh (150 \u0026micro;m) nylon sieve. Approximately 0.3 g (\u0026plusmn;\u0026thinsp;0.1 mg) of the powdered sample was weighed into quartz digestion vessels and treated with 8 mL of concentrated nitric acid (65%). Digestion was carried out using a super microwave digestion system (SUPEC EXPEC 790S, Expec-Tech, China) with the following program: ramping from 25\u0026deg;C to 240\u0026deg;C over 20 minutes, holding at 240\u0026deg;C for 30 minutes, then cooling to 50\u0026deg;C. Digested samples were filtered through 0.45 \u0026micro;m PTFE membranes, diluted to 50 mL with ultrapure water, and analyzed for Cd and Zn concentrations using flame atomic absorption spectrophotometry (AAS; ICE-3500, Thermo Fisher Scientific, USA).\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e\u003cb\u003eEndogenous Hormones and Flavonoids Analysis\u003c/b\u003e\u003c/h2\u003e \u003cp\u003eQuantification of endogenous hormones and flavonoids followed an adapted LC-MS/MS method referencing protocols by Park et al. (Park et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) and Mustafa et al. (Mustafa et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Analytical standards were sourced from Sigma-Aldrich and Merck (USA), with solvents and reagents, including formic acid, methanol, and acetonitrile, obtained from Fisher Scientific (USA). For extraction, approximately 100 mg of frozen sample was homogenized in 80% methanol, then centrifuged and filtered through a 0.22 \u0026micro;m membrane before LC-MS/MS analysis. A Vanquish UHPLC coupled with a Quantis TSQ mass spectrometer (Thermo Fisher Scientific, USA) facilitated chromatographic separation using a C18 column under gradient conditions (10\u0026ndash;90% acetonitrile containing 0.1% formic acid; flow rate of 0.3 mL/min; temperature at 30\u0026deg;C). Detection was performed using electrospray ionization (ESI) and multiple reaction monitoring (MRM) with instrument settings optimized as detailed in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eMeasurement of Reactive Oxygen Species and Antioxidant Enzyme Activities\u003c/h3\u003e\n\u003cp\u003eHydrogen peroxide (H₂O₂) and superoxide anion (O₂\u0026bull;⁻) levels, as well as the activities of peroxidase (POD), superoxide dismutase (SOD), and catalase (CAT), were determined using commercially available assay kits according to the manufacturers' protocols, with minor modifications to sample extraction and handling. For ROS measurements, approximately 0.1 g of tissue was homogenized in liquid nitrogen, extracted overnight in 80% methanol at 4\u0026deg;C, and centrifuged at 10,000 \u0026times; g for 10 min to obtain the supernatant for analysis. H₂O₂ and O₂\u0026bull;⁻ contents were then quantified spectrophotometrically using the corresponding kits (Solarbio, China). For antioxidant enzyme assays, approximately 0.5 g of tissue was homogenized in chilled phosphate buffer (pH 7.0) and centrifuged at 12,000 \u0026times; g for 20 min to obtain crude enzyme extracts. POD, SOD, and CAT activities were measured using commercial assay kits (Sangon Biotech, China) based on guaiacol oxidation, inhibition of nitroblue tetrazolium photoreduction, and H₂O₂ decomposition, respectively. Absorbance was recorded at 470 nm for POD, 560 nm for SOD, and 240 nm for CAT using a spectrophotometer (Presee T9CS, China). Protein concentration was determined by the Bradford method, and enzyme activities were normalized to protein.\u003c/p\u003e\n\u003ch3\u003eGene Expression Analysis by RT-qPCR\u003c/h3\u003e\n\u003cp\u003eGene expression levels of metal transporter genes in rice were quantified by real-time reverse transcription quantitative PCR (RT-qPCR). Twelve representative cultivars were selected from the 44-cultivar panel based on the Zn–Cd response landscape to cover all four response quadrants and contrasting response types. Root samples from these 12 cultivars were used for transporter-expression analysis to provide targeted support for the multivariate candidate signal structures inferred from the sPLS analysis. Total RNA was extracted from root tissues using a Polysaccharide and Polyphenol Plant RNA Extraction Kit (Servicebio, China). RNA concentration and purity were assessed using a NanoDrop spectrophotometer (Thermo Scientific, USA). First-strand cDNA was synthesized from 1 µg of total RNA using the MightyScript First Strand cDNA Synthesis Master Mix (Sangon Biotech, China) according to the manufacturer's instructions and then diluted tenfold with nuclease-free water. RT-qPCR was performed using the 2× SG Fast qPCR Master Mix (Sangon Biotech, China) on a CFX Opus 96 Real-Time PCR System (Bio-Rad, USA). The rice actin gene (\u003cem\u003eOsACT1\u003c/em\u003e) was used as the reference gene (Liu et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e; Wang et al., \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). Gene-specific primers targeting key Cd- and Zn-related transporter genes are listed in Table S2. Primer specificity was verified by melting curve analysis (65–95°C), and amplification efficiency was determined using standard curves generated from fivefold serial dilutions of cDNA. Relative transcript levels were calculated using the 2\u003csup\u003e−ΔΔCt\u003c/sup\u003e method. Each sample included three biological replicates and three technical replicates.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eData analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses and visualizations were performed in R (version 4.5.3). Pearson correlation coefficients were calculated to assess associations between baseline metal concentrations and Zn-induced physiological response variables.\u003c/p\u003e \u003cp\u003eTo quantify the directional and magnitude diversity of Zn–Cd interactions across cultivars, Log₂ fold changes (Log₂FC) of root and shoot Cd concentrations between Zn + Cd and Cd-only treatments were calculated for all 44 cultivars and used to construct a two-dimensional response landscape. Response intensity was defined as the Euclidean distance of each cultivar's coordinate from the origin, and cultivars were classified into four response quadrants based on the sign of root and shoot Log₂FC.\u003c/p\u003e \u003cp\u003eRandom Forest models were constructed using the ranger package (v0.18.0) with 1,000 trees and permutation-based variable importance. Two classification models (M1 and M2) were fitted to predict the direction of Zn-induced changes in shoot Cd accumulation (LCd_delta class) and root-to-shoot translocation efficiency (TF_delta class), respectively. Model performance was evaluated using out-of-bag (OOB) predictions, and accuracy and balanced accuracy were calculated from OOB confusion matrices. To assess whether classification performance exceeded random expectation, permutation tests were performed for M1 and M2 by repeatedly shuffling class labels (1,000 permutations) and recalculating model performance metrics; empirical one-sided p-values were obtained by comparing the observed accuracy, balanced accuracy, AUC, and kappa values against the corresponding null distributions. Two regression models (M3 and M4) were fitted to predict the magnitude of Zn-induced changes in total Cd accumulation (ΔTotalCd) and translocation efficiency (ΔTF), respectively, with observed-versus-predicted R², RMSE, and MAE used to summarize model performance. For all models, predictor variables comprised baseline metal-status traits (leaf and root Cd and Zn concentrations under Cd-only conditions) and Zn-induced physiological response variables (Z − C differences in growth and photochemical parameters). The mtry parameter was set to floor(\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sqrt{p}\\)\u003c/span\u003e\u003c/span\u003e) for classification models and floor(p/3) for regression models, where p denotes the number of predictor variables. Two additional sensitivity models (M1-alt and M2-alt) were fitted using absolute endpoint values as response variables to assess whether the prominence of baseline leaf Cd concentration reflected arithmetic coupling rather than an independent biological signal. Following the Random Forest analysis, two-way ANOVA was additionally applied using base R functions to a targeted subset of core metal-status, growth, and photochemical traits to assess whether these variables also showed significant cultivar, treatment, and cultivar × treatment effects from a complementary univariate, design-based perspective. P values were adjusted using the Benjamini–Hochberg method within each effect category. The RF layer was used to identify informative cultivar-level predictors associated with response direction or magnitude at the population scale. In contrast, the downstream sPLS layer was applied to the representative 12-cultivar subset to resolve deeper multivariable structures underlying specific response dimensions.\u003c/p\u003e \u003cp\u003eSparse partial least squares (sPLS) regression was performed using the mixOmics package (v6.34.0) on a subset of 12 representative cultivars selected from the 44-cultivar panel to cover all four Zn–Cd response quadrants and contrasting response types. Three models were constructed to represent distinct response layers: S1 for Zn-induced changes in translocation efficiency (TF_delta), S2 for changes in leaf Cd accumulation (LCd_delta), and S3 for changes in total Cd accumulation (totalCd_delta). The predictor matrix comprised Z − C delta values of a mechanistically enriched variable set encompassing ROS/antioxidant traits, HMA transporter expression, phenolics/flavonoids, photosynthetic pigments, and hormones. All predictor variables were centered and scaled before model fitting. Optimal numbers of components and retained variables per component were selected via leave-one-out cross-validation using the tune.spls function with Pearson correlation as the optimization criterion; final models were fitted with two components and five retained variables per component. For structural interpretation, retained-variable identities and loading magnitudes were extracted from each component of the three final sPLS models and compared across models. Variables retained uniquely in one response layer, or repeatedly contributing to one model but not the others, were used to describe layer-enriched candidate structures and to assess whether TF_delta-, LCd_delta-, and totalCd_delta-associated signal spaces showed substantial overlap or relative partitioning at the retained-variable level. Given the limited sample size and the absence of usable Q² output under the current cross-validation setting, sPLS models were interpreted as candidate-signal extraction tools rather than generalized predictive frameworks, and the cross-model comparison was treated descriptively rather than as formal evidence of mechanistic independence.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e \u003cb\u003eBaseline Metal Loading Is Associated with the Directional and Physiological Response of Rice to Zn–Cd Interaction: A Quantitative Landscape Analysis\u003c/b\u003e \u003c/p\u003e\u003cp\u003eTo quantify the directional and magnitude diversity of Zn–Cd interactions across rice genotypes, Log₂ fold changes (Log₂FC) of Cd concentrations in roots and shoots between Cd-only and Zn + Cd treatments were calculated for 44 cultivars and visualized as a response landscape (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea). Overall response intensity, defined as the Euclidean distance of each cultivar's coordinate from the origin, ranged from 0.02 to 3.89 (mean = 1.05), indicating substantial genotypic variation in response to the tested Zn co-exposure condition. Based on the sign of root and shoot Log₂FC, cultivars were classified into four response modes: Dual Inhibition (n = 14), in which both root and shoot Cd declined simultaneously; Shoot Promotion/Root Inhibition (n = 13), characterized by decreased root Cd alongside increased shoot Cd; Root Promotion/Shoot Inhibition (n = 11), showing the opposite pattern; and Dual Promotion (n = 6), in which Cd increased in both organs. Cultivar 7 exhibited the highest overall intensity (3.89), driven primarily by strong shoot suppression (shoot Log₂FC = − 3.88), whereas the root response was comparatively modest (root Log₂FC = − 0.28). Cultivar 43 (intensity = 2.87) occupied the upper-left extreme of the Shoot Promotion quadrant, while cultivar 27 was the most distal representative of the Dual Promotion quadrant.\u003c/p\u003e\u003cp\u003eBeyond Cd accumulation, under the tested Zn + Cd co-exposure condition, Zn addition was accompanied by widespread yet highly divergent physiological responses across all 44 cultivars (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb). Among photosystem indicators, Y(II) increased in 30 out of 44 cultivars (68.2%; median Δ = +0.016), and ETR showed a similar trend (65.9% positive; median Δ = +3.05), suggesting a general tendency toward improved photochemical efficiency, albeit with considerable inter-cultivar variation. In contrast, qP decreased in the majority of cultivars (27/44, 61.4%; median Δ = −0.037), indicating that photochemical quenching capacity responded more heterogeneously across genotypes. For growth-related traits, root fresh weight declined in 31 of 44 cultivars (70.5%; median Δ = −0.25 g), representing the most consistent directional response across the dataset, while shoot fresh weight and total biomass also showed predominantly negative shifts (61.4% and 63.6%, respectively). Shoot water content was a notable exception, increasing in 63.6% of cultivars (median Δ = +2.80%). Root activity also showed substantial genotype-dependent variability (SD = 7.24). Although 61.4% of cultivars exhibited a positive response, with a median Δ of + 2.37, the mean remained slightly negative (− 0.36), indicating that a subset of cultivars experienced pronounced declines that offset the more common moderate increases. Collectively, these patterns indicate that Zn addition triggers highly genotype-specific physiological shifts across photosystem efficiency, biomass allocation, root activity, and water status, thereby defining a multidimensional response background against which Zn–Cd interaction types can be further resolved by random forest analysis.\u003c/p\u003e\u003cp\u003e(a) Log₂ fold change (Log₂FC) of root and shoot Cd concentrations between Zn + Cd and Cd-only treatments, plotted as a two-dimensional response landscape for 44 rice cultivars. Each point represents one cultivar; point size and color indicate response intensity (Euclidean distance from the origin: Low, Medium, High). Dashed lines demarcate four response quadrants: Dual Promotion (root↑ shoot↑, n = 6), Shoot Promotion/Root Inhibition (root↓ shoot↑, n = 13), Dual Inhibition (root↓ shoot↓, n = 14), and Root Promotion/Shoot Inhibition (root↑ shoot↓, n = 11). Selected cultivar IDs are labeled. (b) Strip plots showing the distribution of Δ values (Zn + Cd minus Cd-only) for 11 physiological variables across all 44 cultivars. Each dot represents one cultivar; red = positive response (increase), blue = negative response (decrease). The solid vertical line in each panel indicates the median; the dashed line marks zero (no change). Panels are grouped by module: photosystem indicators (Fv/Fm, Y(II), ETR, qP, Y(NPQ); blue background), root physiology (Root Activity; green background), and growth traits (Biomass, Root FW, Shoot FW, Root WC, Shoot WC; orange background).\u003c/p\u003e\u003cp\u003e \u003cb\u003eRandom Forest analysis identifies cultivar-specific Zn–Cd response types and their associated predictors.\u003c/b\u003e \u003c/p\u003e\u003cp\u003eTo dissect the factors underlying the divergent Zn–Cd interaction patterns across cultivars, we constructed four primary Random Forest (RF) models that address both response direction and magnitude. M1 and M2 were classification models used to predict whether Zn induced a promoted or inhibited response in shoot Cd accumulation and root-to-shoot translocation, respectively, with M2 serving as the primary model for translocation-related response types. M3 and M4 were regression models used to quantify the magnitude of Zn-mediated changes in total Cd accumulation (Δ𝑇𝑜𝑡𝑎𝑙𝐶𝑑) and translocation efficiency (Δ𝑇𝐹). In all models, predictors were structured into two layers: baseline metal-status traits, representing the pre-existing elemental chassis, and physiological response traits, representing Zn-induced functional shifts. To assess whether the arithmetic relationship between the baseline trait L Cd Conc_C and the delta-based response variables inflated their variable importance rankings, two sensitivity models (M1-alt and M2-alt) were additionally fitted using absolute endpoint values (L Cd Conc_Z and TF_Z, respectively) as response variables, with identical predictor sets. This design provided a common framework for evaluating the joint contribution of elemental background and physiological response to cultivar-specific Zn–Cd interaction phenotypes.\u003c/p\u003e\u003cp\u003eModel performance evaluation revealed a clear contrast between the predictability of response direction and that of response magnitude. The two classification models showed moderate but meaningful discriminative capacity, with M1 (shoot Cd response) and M2 (translocation response) reaching accuracies of 70.45% and 68.18%, and balanced accuracies of 68.32% and 67.91%, respectively. By contrast, the two regression models explained only 13.79% (M3) and 23.82% (M4) of the variance. This overall pattern indicates that the Zn-induced Cd response was more readily classified in terms of direction than quantitatively predicted in terms of magnitude. Permutation tests further indicated that the observed classification performance was unlikely to be entirely due to random label structure. For M1, both accuracy (0.6818 vs. null mean 0.4954, permutation p = 0.032) and balanced accuracy (0.6632 vs. null mean 0.4744, permutation p = 0.030) exceeded random expectation. For M2, balanced accuracy also marginally exceeded the null distribution (0.6553 vs. null mean 0.4781, permutation p = 0.050). AUC-based comparisons, however, were weaker for both models and did not clearly exceed the corresponding null distributions. Together, these results support the interpretation that the RF models captured modest but non-random population-level structure in response-direction classification, particularly for shoot Cd response direction.\u003c/p\u003e\u003cp\u003eVariable-importance analysis revealed distinct but complementary patterns across the four models. In M1 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ea), L Cd Conc_C ranked first (0.0359), followed by Biomass (Z-C) and L Zn Conc_C, with Y(II) (Z-C) and Fv/Fm (Z-C) also retained among the top five. M2 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ec) showed a closely parallel structure, with L Cd Conc_C again ranking first (0.0423) and L Zn Conc_C, Y(II) (Z-C), Biomass (Z-C), and No. plants (Z-C) in the following order. The convergence of predictor structures between M1 and M2 suggests that the factors associated with shoot Cd accumulation direction and translocation response direction were largely shared across cultivars, consistent with translocation being a primary determinant of shoot Cd accumulation under the present experimental conditions. Of note, L Cd Conc_C ranked first in both models. However, because it contributes arithmetically to both delta-based response variables, its prominence may partly reflect structural coupling rather than fully independent biological information. Sensitivity models showed that removing this coupling did not disrupt the broader ranking structure, but redistributed importance toward different subsets of predictors (Fig. \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003e). Although the sensitivity models showed lower overall predictive performance, consistent with the greater background variance inherent in absolute endpoint values (M1-alt OOB R² = 0.061; M2-alt OOB R² = 0.106; Fig. \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003ea-b), the consistency of key predictors across primary and sensitivity models supports the robustness of the observed importance patterns. In M1-alt (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eb), Biomass (Z-C) and Fv/Fm (Z-C) emerged as the leading predictors, suggesting that cultivar-specific physiological responses to Zn were informative for classifying whether shoot Cd accumulation was promoted or inhibited. In M2-alt (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ed), L Zn Conc_C retained high importance alongside several root metal-status traits, and the shift from a predominantly leaf-based predictor structure in M2 to a more root-enriched structure in M2-alt suggests that the direction of translocation response and the absolute efficiency of root-to-shoot Cd transfer were associated with partially distinct organ-level predictor structures. In the regression models, L Zn Conc_C and R Zn Conc_C were the strongest predictors of translocation magnitude in M4 (Fig. S2c), suggesting that the extent of Zn-induced changes in translocation efficiency was closely associated with the pre-existing Zn accumulation status of individual cultivars. In M3 (Fig. S2a), physiological response variables contributed relatively little, suggesting that prediction of Zn-induced changes in total Cd uptake relied more strongly on baseline metal-status traits than on the physiological response variables included here. Together, these patterns indicate that baseline elemental status and Zn-induced physiological shifts jointly contributed to cultivar-specific Zn–Cd interaction phenotypes, but with stronger relevance for response classification than for quantitative prediction of response magnitude.\u003c/p\u003e\u003cp\u003eThe regression models showed weaker overall explanatory performance. The low predictive power of both M3 and M4 indicates that the absolute magnitude of the Zn–Cd interaction could not be robustly captured by the current variable set alone, and several cultivars showed large residuals in M4, indicating substantial model-unexplained heterogeneity among genotypes. Together, these results suggest that the variables used here were more effective in discriminating response type than in quantitatively resolving response intensity.\u003c/p\u003e\u003cp\u003eHorizontal bars represent permutation importance values for each predictor variable. (a) M1: classification model for shoot Cd response direction (response = LCd_delta class; n = 44). (b) M1-alt: sensitivity model using absolute shoot Cd concentration as the response variable (response = L Cd Conc_Z; n = 44). (c) M2: classification model for translocation response direction (response = TF_delta class; n = 44). (d) M2-alt: sensitivity model using absolute translocation factor as the response variable (response = TF_Z; n = 44). All four models used identical predictor sets comprising baseline metal-status traits (L/R Cd and Zn concentrations under control conditions) and Zn-induced physiological response traits (Z − C differences in growth and photochemical parameters). Negative importance values indicate that including the variable degraded model predictions. M1-alt and M2-alt were fitted to assess whether the prominence of L Cd Conc_C in M1 and M2 reflected arithmetic coupling with the delta-based response variables rather than an independent biological signal. Complementary univariate analyses further supported the RF framework. Across the selected follow-up traits, two-way ANOVA consistently detected significant cultivar effects and cultivar × treatment interactions after BH correction (Table S3), indicating that Zn-induced variation in metal status, growth, and photochemical traits was strongly genotype-dependent. Treatment main effects were also significant for most variables except leaf Cd concentration.\u003c/p\u003e\u003ch2\u003esPLS analysis reveals layer-specific multivariable structures underlying Zn-induced Cd response variation\u003c/h2\u003e\u003cp\u003eTo further resolve the deep-layer signals associated with cultivar-specific Zn–Cd interaction patterns, we performed sparse partial least squares (sPLS) analyses using the 12 selected cultivars and a mechanistically enriched variable set, including ROS/antioxidant traits, \u003cem\u003eHMA\u003c/em\u003e transporter expression, phenolics/flavonoids, pigments, and hormones. Three models were constructed to represent different response layers: S1 for Zn-induced changes in translocation efficiency (TF_delta), S2 for changes in leaf Cd accumulation (LCd_delta), and S3 for changes in total Cd accumulation (totalCd_delta). All models were fitted with two components and five retained variables per component, yielding similarly high fitting performance (R² = 0.904, 0.789, and 0.872 for S1, S2, and S3, respectively; observed-versus-predicted plots shown in Figure S3). However, given the limited sample size and absence of Q² values, these models were interpreted primarily as tools for candidate-signal extraction rather than generalized prediction. Accordingly, the 12-cultivar sPLS analysis was positioned as a deeper interpretive layer that complemented rather than replaced the 44-cultivar Random Forest framework.\u003c/p\u003e\u003cp\u003eAmong the three models, S1 showed the most integrative structure for Cd translocation response (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea, b). Its retained variables spanned two distinct dimensions. Comp1 was dominated by R_O\u003csub\u003e2\u003c/sub\u003e•\u003csup\u003e−\u003c/sup\u003e_delta (loading = -0.778), with R_Astragalin_delta (loading = -0.606) providing additional phenolic/flavonoid-related structure, and R_HMA2_delta, L_Astragalin_delta, and R_POD_delta contributing smaller supplementary signals. This primary axis indicated that Zn-induced root superoxide and flavonoid responses were the major features associated with separation along the primary translocation-related latent dimension. Comp2 revealed a secondary dimension dominated by R_SA_delta (loading = -0.831), with R_\u003cem\u003eHMA1\u003c/em\u003e_delta (loading = 0.467) and L_O\u003csub\u003e2\u003c/sub\u003e•\u003csup\u003e−\u003c/sup\u003e_delta (loading = 0.256) contributing additional transporter- and antioxidant-associated structure, indicating that salicylate-related reconfiguration defined a secondary hormone-associated axis within the translocation layer.\u003c/p\u003e\u003cp\u003eS2 revealed a different pattern for Zn-induced changes in leaf Cd accumulation (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ec, d). Comp1 was driven primarily by flavonoid- and hormone-related variables, with L_Quercetin_delta showing the strongest loading (0.734) and R_JA_delta contributing substantial additional support (loading = 0.652); L_SL_delta, L_Cinnamic_Acid_delta, and R_\u003cem\u003eHMA1\u003c/em\u003e_delta were retained with smaller positive loadings. Comp2 defined a distinct secondary axis dominated by R_Astragalin_delta (loading = -0.752), with L_Protamine_Sulfate_delta (loading = 0.480) and Carotenoid_delta (loading = -0.439) as the next strongest contributors, and L_SOD_delta and L_H\u003csub\u003e2\u003c/sub\u003eO\u003csub\u003e2\u003c/sub\u003e_delta providing minor antioxidant-associated signals. Together, the two components indicated that leaf quercetin/jasmonate responses defined the primary leaf Cd accumulation axis, whereas a secondary dimension captured root astragalin and leaf carotenoid co-variation.\u003c/p\u003e\u003cp\u003eS3, in turn, showed a third-variable structure in total Cd response amplitude (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ee, f). Comp1 was overwhelmingly dominated by R_Quinic_Acid_delta (loading = 0.928), with R_Quercetin_delta retained at a moderate negative loading (-0.363) and L_Rosmarinic_Acid_delta, L_Quercetin_delta, and R_CAT_delta contributing minor supplementary signals. This primary axis indicated that root quinic acid accumulation was by far the strongest feature associated with total Cd response amplitude. Comp2 revealed a qualitatively distinct secondary dimension centered on HMA transporter expression, with R_\u003cem\u003eHMA7\u003c/em\u003e_delta as the dominant variable (loading = 0.890) and R_HMA3_delta showing a moderate negative loading (-0.403); R_JA_delta and R_CAT_delta contributed smaller hormone- and antioxidant-associated signals. This pattern suggests that while phenolic reconfiguration defined the primary axis of total Cd response variation, a secondary \u003cem\u003eHMA7\u003c/em\u003e/\u003cem\u003eHMA3\u003c/em\u003e-associated transporter dimension was also captured within the same model.\u003c/p\u003e\u003cp\u003eViewed together, the three sPLS models revealed related but non-identical multivariable structures across response layers. At the comp1 level, the dominant retained variables showed clear layer-specific partitioning: S1 was centered on a root superoxide/flavonoid-associated structure (R_O\u003csub\u003e2\u003c/sub\u003e•\u003csup\u003e−\u003c/sup\u003e_delta, R_Astragalin_delta), S2 on a quercetin/jasmonate-associated structure (L_Quercetin_delta, R_JA_delta), and S3 on a root quinic acid-dominated phenolic structure (R_Quinic_Acid_delta). At the comp2 level, each model captured a qualitatively distinct secondary dimension: a salicylate/\u003cem\u003eHMA\u003c/em\u003e-associated axis in S1, an astragalin/carotenoid-associated axis in S2, and an \u003cem\u003eHMA7\u003c/em\u003e/\u003cem\u003eHMA3\u003c/em\u003e transporter axis in S3. Notably, R_O\u003csub\u003e2\u003c/sub\u003e•\u003csup\u003e−\u003c/sup\u003e_delta was dominant in S1 comp1 but absent from S2 and S3 comp1, while HMA7-associated signals appeared as a secondary dimension in S3 but not as a primary feature in either S1 or S2. This limited overlap across both components indicates that Zn-induced variation in translocation efficiency, leaf Cd accumulation, and total Cd accumulation was associated with partially distinct candidate structures rather than a single shared latent mechanism.\u003c/p\u003e\u003cp\u003e(a) Sample scores of S1, the sPLS model for Zn-induced changes in translocation efficiency (TF_delta). (b) Retained variables in S1. (c) Sample scores of S2, the sPLS model for Zn-induced changes in leaf Cd accumulation (LCd_delta). (d) Retained variables in S2. (e) Sample scores of S3, the sPLS model for Zn-induced changes in total Cd accumulation (totalCd_delta). (f) Retained variables in S3. Across the three models, the retained-variable structures were partially overlapping but non-identical. S1 comp1 was dominated by root superoxide and flavonoid signals (R_O\u003csub\u003e2\u003c/sub\u003e•\u003csup\u003e−\u003c/sup\u003e_delta, R_Astragalin_delta), with a secondary salicylate/HMA-associated axis (R_SA_delta, R_\u003cem\u003eHMA1\u003c/em\u003e_delta) in comp2. S2 comp1 was defined by leaf quercetin and jasmonate responses (L_Quercetin_delta, R_JA_delta), with a secondary astragalin/carotenoid axis (R_Astragalin_delta, Carotenoid_delta) in comp2. S3 comp1 was overwhelmingly dominated by root quinic acid (R_Quinic_Acid_delta), with a secondary HMA transporter-associated axis (R_HMA7_delta, R_\u003cem\u003eHMA3\u003c/em\u003e_delta) in comp2.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe present study indicates that Zn-Cd interaction in rice cannot be described adequately as a uniformly antagonistic process under the tested near-equimolar co-exposure regime (Cai et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Across 44 cultivars, the response landscape revealed a four-quadrant distribution of interaction types and a nearly 200-fold range in response intensity, showing that Zn-associated changes in Cd accumulation varied substantially among genotypes (Tavarez et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Beyond Cd redistribution, Zn addition was accompanied by divergent shifts in photosystem efficiency, biomass allocation, root activity, and water status, indicating that the observed heterogeneity spans multiple physiological dimensions rather than a single trait (Chang et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This heterogeneity was not random with respect to the measured predictor set: Random Forest models classified response direction with approximately 70% accuracy using baseline metal status and Zn-induced physiological shifts (Wright and Ziegler, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). These results do not establish causality, but they do indicate that response direction is associated with partially discriminable biological features captured by the present framework. Taken together, the data support treating Zn-Cd interaction outcome as a structured, genotype-associated response phenotype and justify analyzing response direction and response intensity as related but partially distinct dimensions rather than collapsing them into a single response metric (Mashabela et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBeyond discriminating response direction, the RF models revealed several coarse-grained interpretive patterns with independent analytical value. First, the near-identical predictor structures of M1 and M2\u0026mdash;with L Cd Conc_C ranking first in both models and Zn-induced physiological shifts (Biomass Z-C, Y(II) Z-C, Fv/Fm Z-C) consistently retained among the top predictors\u0026mdash;indicate that shoot Cd accumulation direction and translocation response direction are associated with a largely shared set of cultivar-level features, suggesting that the genotypic features associated with whether Zn promotes or inhibits root-to-shoot translocation substantially overlap with those associated with shoot Cd accumulation direction\u0026mdash;consistent with translocation being the dominant process through which genotypic identity may shape shoot Cd outcomes, with little evidence of systematic decoupling between translocation and leaf accumulation responses across the cultivar panel (Uraguchi et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Second, variable importance analysis in M4 revealed that baseline Zn status in both leaves and roots (L Zn Conc_C, importance\u0026thinsp;=\u0026thinsp;0.659; R Zn Conc_C, importance\u0026thinsp;=\u0026thinsp;0.536) was the dominant predictor of translocation magnitude, substantially outranking all physiological response variables, several of which contributed negatively (Fv/Fm Z-C: \u0026minus;0.054; Biomass Z-C: \u0026minus;0.136). This pattern indicates that the extent of Zn-induced translocation shifts is more closely tied to a cultivar's pre-existing Zn accumulation status than to its observable physiological response trajectory. We therefore infer that baseline Zn homeostatic status may be associated with translocation response amplitude, possibly reflecting differences in cellular redox and metal-sensing states relevant to downstream transporter responses\u0026mdash;a candidate hypothesis examined in the sPLS analysis below. Third, the shift from a predominantly leaf-based predictor structure in M2 to a more root-enriched structure in M2-alt suggests that the directional and quantitative dimensions of translocation response are associated with partially distinct organ-level informational layers\u0026mdash;with root metal status more closely linked to absolute translocation capacity (Uraguchi et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) and leaf physiological variables more informative for classifying response direction, possibly reflecting systemic differentiation between shoot and root response layers. Whether this organ-level asymmetry reflects shoot-to-root signaling, shared genotypic architecture expressed differently across organs, or both, remains to be determined, and is examined at the candidate-signal level in the sPLS analysis below. Permutation-based null comparisons further supported the interpretation that the RF models captured non-random discriminative structure rather than purely chance separation. However, this signal was stronger for shoot Cd response direction than for translocation response direction, and remained modest in probability-ranking terms as reflected by the weaker AUC-based comparisons.\u003c/p\u003e \u003cp\u003eA key interpretive implication of the three-tier framework concerns the relationship between the RF-detectable surface signal and the sPLS-resolved candidate structure. The RF layer captured cultivar-level predictors that were informative for response classification, but did not by itself determine whether photochemical, transporter-related, or redox-associated variables should be regarded as primary or secondary within a given response layer (Mashabela et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The sPLS comparison helped reduce this ambiguity by showing that the apparently informative physiological signal did not behave as a single shared mechanistic axis across all response dimensions (Chun and Keleş, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Instead, translocation-related variation was associated mainly with a root superoxide/flavonoid structure, with R_O\u003csub\u003e2\u003c/sub\u003e\u0026bull;\u003csup\u003e\u0026minus;\u003c/sup\u003e_delta and R_Astragalin_delta forming the dominant comp1 variables and additional comp2 contributions from salicylate- and transporter-related variables (Kostyuk et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2004\u003c/span\u003e), whereas leaf Cd accumulation was associated more closely with a quercetin/jasmonate co-varying structure, with L_Quercetin_delta and R_JA_delta dominating comp1 and additional comp2 contributions from astragalin- and carotenoid-related variables (Lei et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). This pattern is consistent with the interpretation that flavonoid and hormonal signals function as layer-dependent phenotypic readouts of deeper metal-homeostasis/redox states rather than as a single shared upstream driver (Considine and Foyer, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTaken together, these RF-derived patterns highlight three coarse-grained population-level inferences: translocation appears to be the primary axis associated with shoot Cd response direction; baseline Zn homeostasis is associated with translocation response magnitude; and root- and leaf-derived predictor layers were differentially associated with translocation direction versus efficiency. However, the regression models also revealed a clear resolution limit: M4 explained only 23.8% of the variance in translocation magnitude, and M3 explained only 13.8% of the variance in total Cd uptake. The genotypes with the most extreme translocation responses were precisely those least captured by the present variable set, suggesting that intensity outliers may harbor biological features qualitatively absent from the surface-level predictor space. The limited predictive power of the regression models indicates that RF analysis across 44 cultivars functioned primarily as a coarse-grained filter, sufficient to capture the directional patterning of Zn\u0026ndash;Cd interaction phenotypes but unable to resolve the finer-grained biological variation underlying response intensity. These follow-up univariate results reinforce the interpretation that Zn\u0026ndash;Cd response diversity is genotype-dependent not only at the multivariate pattern-recognition level captured by RF, but also at the level of conventional trait-by-treatment interaction. The variables associated with response intensity\u0026mdash;including HMA transporter expression, ROS/antioxidant status, hormonal signaling, and secondary metabolite profiles\u0026mdash;operate at a deeper biological resolution than can be captured by population-scale physiological screening alone. The sPLS analysis on the 12 representative cultivars was therefore designed as a fine-grained filtering step operating on a qualitatively distinct variable set, thereby extending the analysis into a deeper candidate-signal space encompassing HMA transporter expression, ROS/antioxidant status, hormonal signaling, and secondary metabolite profiles, and identifying layer-specific candidate signal structures as testable hypotheses.\u003c/p\u003e \u003cp\u003eThe sPLS analysis conducted on the 12 representative cultivars indicated that Zn\u0026ndash;Cd interaction diversity is multivariate and response-layer specific: no single variable dominated across all three models, and the retained signal structures of S1 (translocation efficiency, TF_delta), S2 (leaf Cd accumulation, LCd_delta), and S3 (total Cd accumulation, totalCd_delta) showed limited overlap at the dominant-variable level. Rather than converging on a single shared latent structure, the three response layers were associated with distinct candidate axes, supporting the view that Zn-induced changes in translocation, leaf accumulation, and total accumulation are organized by partially separable biological structures (Mashabela et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAt the translocation layer (S1), the dominant retained variables point to a root redox/flavonoid-associated structure rather than to a chlorophyll- or pigment-dominated one. Comp1 was dominated by R_O\u003csub\u003e2\u003c/sub\u003e\u0026bull;\u003csup\u003e\u0026minus;\u003c/sup\u003e_delta (loading = -0.778) and R_Astragalin_delta (loading = -0.606), indicating that Zn-induced root superoxide reconfiguration and astragalin-related flavonoid shifts were the most prominent features associated with the primary translocation axis in the present dataset (Feigl et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Tian et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). A secondary dimension represented by comp2 additionally retained strong salicylate-related variation, with R_SA_delta showing the largest loading, while HMA-related transporter signals, including R_\u003cem\u003eHMA2\u003c/em\u003e_delta, contributed further transporter-associated structure(Kim and Jang, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Lee et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Sasaki et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Ueno et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Because these relationships derive from retained-variable composition and loading structure rather than direct perturbation, they are best interpreted as candidate features of a translocation-associated redox/flavonoid axis rather than as proof of a causal sequence. Integrating the S1 observations, we propose a candidate redox-baseline framework in which cultivar-specific root redox state under Cd-alone conditions may influence the amplitude of Zn-induced redox reorganization and thereby contribute to variation in Zn-Cd interaction direction (Cai et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Umair Hassan et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). In this view, redox and flavonoid status would act not as isolated explanatory variables, but as parts of a broader translocation-associated candidate structure that also includes salicylate- and HMA-related transporter responses. This interpretation remains provisional and should be tested in broader germplasm panels and targeted perturbation experiments.\u003c/p\u003e \u003cp\u003eAt the leaf Cd accumulation layer (S2), the predictor structure shifted toward a quercetin/jasmonate-associated organization. Within the current retained-variable structure, comp1 was dominated by L_Quercetin_delta (loading\u0026thinsp;=\u0026thinsp;0.734) and R_JA_delta (loading\u0026thinsp;=\u0026thinsp;0.652), with L_SL_delta and L_Cinnamic_Acid_delta providing additional phenolic/hormonal support, indicating that Zn-induced leaf flavonoid accumulation and root jasmonate signaling were the primary features associated with the leaf Cd accumulation axis (Chen et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). A secondary dimension, represented by comp2, additionally retained R_Astragalin_delta and Carotenoid_delta, along with minor antioxidant-associated signals, suggesting that a partially distinct astragalin/carotenoid co-varying dimension also contributed to the leaf accumulation layer (Ramel et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Relative to S1, this pattern indicates that leaf Cd accumulation was associated less with the root superoxide-dominated structure characterizing the translocation layer, and more with a flavonoid/hormone-centered axis in which quercetin and jasmonate responses co-varied with secondary phenolic and pigment shifts (B. Wang et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The present data support association, not directional causation, and therefore do not establish whether the quercetin- and jasmonate-related signals are upstream regulators of Cd partitioning, downstream stress readouts, or parallel components of a broader leaf-level response structure.\u003c/p\u003e \u003cp\u003eThe difference between S1 and S2 is important because it suggests that the physiological signals retained within the broader analytical framework should not be interpreted as a single universal mechanism across all Zn-Cd response dimensions (Mashabela et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Instead, their interpretive meaning appears to depend on which response layer is being examined. In the present dataset, the strongest translocation-associated signal was linked to a root superoxide/flavonoid structure centered on R_O\u003csub\u003e2\u003c/sub\u003e\u0026bull;\u003csup\u003e\u0026minus;\u003c/sup\u003e_delta and R_Astragalin_delta (Zhang et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). In contrast, leaf Cd accumulation aligned more closely with a quercetin/jasmonate co-varying structure, where flavonoid and hormonal shifts likely reflect leaf-level stress responses accompanying Cd partitioning rather than serving as direct Cd transport routes (Lei et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). This layer-specific partitioning strengthens the view that RF-level physiological signals are better interpreted as entry points into deeper, layer-specific candidate structures than as self-sufficient mechanistic explanations.\u003c/p\u003e \u003cp\u003eThe present dataset also suggests broader implications for the evaluation of Zn-based mitigation strategies in rice. Because Zn was used here as a model co-exposure factor rather than a field-optimized agronomic amendment, the present results do not justify general claims about the universal safety or efficacy of Zn supplementation (Tavarez et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; S. Wang et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). They do, however, indicate that cultivar background may condition whether Zn co-exposure is associated with inhibition, promotion, or more complex redistribution of Cd under a fixed near-equimolar regime (Cai et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Tavarez et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This raises the possibility that genotype-dependent responsiveness may represent a broader candidate dimension in rice Cd-risk evaluation. However, that broader applicability remains to be tested before any screening or management application is claimed.\u003c/p\u003e \u003cp\u003eAn additional implication of this framework is that constitutive low-Cd tendency and co-exposure responsiveness may not fully overlap as cultivar dimensions (Sasaki et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Tavarez et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Traits that reduce constitutive Cd entry or retention (Sasaki et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Ueno et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) would be expected to lower oxidative burden under Cd stress, whereas the present data suggest that a stronger root redox baseline state may be associated with a more responsive translocation-related layer under Zn-Cd co-exposure (Uraguchi et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Xue et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Although this possibility is not tested directly here, it raises the question of whether cultivars optimized for low constitutive Cd accumulation and those showing stronger responsiveness to external modulation necessarily represent the same physiological strategy.\u003c/p\u003e \u003cp\u003eThe layer-specific signal structures identified across S1, S2, and S3 also raise the question of whether the organ-level informational asymmetry detected in the RF models - where leaf baseline traits dominated some coarse-grained classifications, yet root redox or phenolic variables dominated several sPLS latent axes - reflects a general feature of genotype-dependent Zn\u0026ndash;Cd interaction or only the specific multivariate composition of the present cultivar subset (Tavarez et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Uraguchi et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). At a minimum, the current results show that different response layers need not converge on the same dominant retained variables. That apparent cross-organ inconsistency at the RF layer can be reconciled by a deeper candidate-structure view in which different organs contribute differently to different response dimensions(Chun and Keleş, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Mashabela et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eNotwithstanding these layer-specific patterns, a subset of cultivars \u0026mdash; including cultivars 27, 43, 10, and 11 \u0026mdash; repeatedly deviated from the dominant latent structures across one or more sPLS models, as reflected in their sample score distributions and observed-versus-predicted residuals. This residual heterogeneity indicates that the candidate signal profiles identified here, while capturing the dominant axes of variation among the 12 cultivars, do not fully account for the biological complexity present in a subset of genotypes. Cultivars 27 and 43, in particular, showed hormone-rich profiles with coordinated \u003cem\u003eHMA\u003c/em\u003e and antioxidant responses that partially overlapped across multiple signal layers, suggesting that these genotypes may integrate multiple regulatory modules more synergistically than the layer-specific sPLS structures can resolve individually. It should be noted that the sPLS models were interpreted as candidate signal-extraction tools rather than generalized predictive frameworks, given the limited sample size of 12 cultivars and the absence of usable cross-validation Q\u0026sup2; values. The candidate signal patterns identified here, therefore, require validation in larger, more diverse germplasm panels before their generalizability can be established.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, under a fixed near-equimolar Zn\u0026thinsp;+\u0026thinsp;Cd co-exposure condition, rice cultivars exhibited substantial diversity in both the direction and magnitude of Zn-associated changes in Cd accumulation. By integrating response-landscape analysis, Random Forest classification, and sPLS-based candidate-structure extraction, this study shows that response direction was more tractable than response magnitude, and that distinct Zn\u0026ndash;Cd response layers were associated with partially differentiated multivariable structures. Translocation-related variation was linked primarily to a root superoxide/flavonoid-associated structure, leaf Cd accumulation to a more flavonoid/hormone-centered candidate structure, and total Cd amplitude to a phenolic/flavonoid-associated structure. Together, these results support a layer-specific candidate framework for genotype-dependent Zn\u0026ndash;Cd interaction, rather than a single shared mechanism, and highlight root redox baseline state under Cd stress alone as a potential upstream feature associated with response direction. More broadly, the study suggests that constitutive Cd accumulation tendency and co-exposure responsiveness may represent only partially overlapping cultivar dimensions.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflict of Interest\u003c/h2\u003e \u003cp\u003eOn behalf of all authors, the corresponding author states that there is no conflict of interest.\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eEthics, Consent to Participate, and Consent to Publish declarations\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis research was supported by the Natural Science Foundation of National Natural Science Foundation of China (32201392), and Guangxi Major Science and Technology Program (Guike AA24263045).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eS.C. conceived the study, curated the data, developed the methodology, and drafted the manuscript. B.S. contributed to methodology development and software analysis. F.Z. contributed to the investigation and resources. Y.L. contributed to data curation and visualization. X.C. contributed to visualization. Q.L. contributed to data curation, funding acquisition, project administration, and manuscript review and editing. All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eThe authors would like to thank all who contributed to this work.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e \u003cp\u003eAll data generated or analyzed during this study will be available upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAdil MF, Sehar S, Chen G, Chen Z-H, Jilani G, Chaudhry AN, Shamsi IH (2020) Cadmium-zinc cross-talk delineates toxicity tolerance in rice via differential genes expression and physiological / ultrastructural adjustments. Ecotoxicol Environ Saf 190:110076. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ecoenv.2019.110076\u003c/span\u003e\u003cspan address=\"10.1016/j.ecoenv.2019.110076\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai Y, Xu W, Wang M, Chen W, Li X, Li Y, Cai Y (2019) Mechanisms and uncertainties of Zn supply on regulating rice Cd uptake. Environ Pollut Barking Essex 1987 253:959\u0026ndash;965. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envpol.2019.07.077\u003c/span\u003e\u003cspan address=\"10.1016/j.envpol.2019.07.077\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang W, Wang W, Shi Z, Cao G, Zhao X, Su X, Chen Y, Wu J, Yang Z, Liu C, Shang L, Cai Z (2023) Comparative metabolomics combined with physiological analysis revealed cadmium tolerance mechanism in indica rice (\u003cem\u003eOryza sativa\u003c/em\u003e L). J Agric Food Chem 71:7669\u0026ndash;7678. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.jafc.3c00850\u003c/span\u003e\u003cspan address=\"10.1021/acs.jafc.3c00850\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen X, Jiang W, Tong T, Chen G, Zeng F, Jang S, Gao W, Li Z, Mak M, Deng F, Chen Z-H (2021) Molecular interaction and evolution of jasmonate signaling with transport and detoxification of heavy metals and metalloids in plants. Front Plant Sci 12:665842. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpls.2021.665842\u003c/span\u003e\u003cspan address=\"10.3389/fpls.2021.665842\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChun H, Keleş S (2010) Sparse partial least squares regression for simultaneous dimension reduction and variable selection. J R Stat Soc Ser B Stat Methodol 72:3\u0026ndash;25. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1467-9868.2009.00723.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1467-9868.2009.00723.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConsidine MJ, Foyer CH (2014) Redox regulation of plant development. Antioxid Redox Signal 21:1305\u0026ndash;1326. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1089/ars.2013.5665\u003c/span\u003e\u003cspan address=\"10.1089/ars.2013.5665\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDu J, Zeng J, Ming X, He Q, Tao Q, Jiang M, Gao S, Li X, Lei T, Pan Y, Chen Q, Liu S, Yu X (2020) The presence of zinc reduced cadmium uptake and translocation in \u003cem\u003eCosmos bipinnatus\u003c/em\u003e seedlings under cadmium/zinc combined stress. Plant Physiol Biochem 151:223\u0026ndash;232. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.plaphy.2020.03.019\u003c/span\u003e\u003cspan address=\"10.1016/j.plaphy.2020.03.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeigl G, Lehotai N, Moln\u0026aacute;r \u0026Aacute;, \u0026Ouml;rd\u0026ouml;g A, Rodr\u0026iacute;guez-Ruiz M, Palma JM, Corpas FJ, Erdei L, Kolbert Z (2015) Zinc induces distinct changes in the metabolism of reactive oxygen and nitrogen species (ROS and RNS) in the roots of two brassica species with different sensitivity to zinc stress. Ann Bot 116:613\u0026ndash;625. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/aob/mcu246\u003c/span\u003e\u003cspan address=\"10.1093/aob/mcu246\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFontanili L, Lancilli C, Suzui N, Dendena B, Yin Y-G, Ferri A, Ishii S, Kawachi N, Lucchini G, Fujimaki S, Sacchi GA, Nocito FF (2016) Kinetic analysis of zinc/cadmium reciprocal competitions suggests a possible Zn-insensitive pathway for root-to-shoot cadmium translocation in rice. Rice 9:16\u0026ndash;28. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12284-016-0088-3\u003c/span\u003e\u003cspan address=\"10.1186/s12284-016-0088-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaider FU, Liqun C, Coulter JA, Cheema SA, Wu J, Zhang R, Wenjun M, Farooq M (2021) Cadmium toxicity in plants: impacts and remediation strategies. Ecotoxicol Environ Saf 211:111887. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ecoenv.2020.111887\u003c/span\u003e\u003cspan address=\"10.1016/j.ecoenv.2020.111887\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu J (2021) Toward unzipping the ZIP metal transporters: structure, evolution, and implications on drug discovery against cancer. FEBS J 288:5805\u0026ndash;5825. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/febs.15658\u003c/span\u003e\u003cspan address=\"10.1111/febs.15658\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHussain B, Ashraf MN, Shafeeq-Ur-Rahman N, Abbas A, Li J, Farooq M (2021) Cadmium stress in paddy fields: effects of soil conditions and remediation strategies. Sci Total Environ 754:142188. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2020.142188\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2020.142188\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim G-N, Jang H-D (2009) Protective mechanism of quercetin and rutin using glutathione metabolism on HO-induced oxidative stress in HepG2 cells. Ann N Y Acad Sci 1171:530\u0026ndash;537. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1749-6632.2009.04690.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1749-6632.2009.04690.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKostyuk VA, Potapovich AI, Strigunova EN, Kostyuk TV, Afanas'ev IB (2004) Experimental evidence that flavonoid metal complexes may act as mimics of superoxide dismutase. Arch Biochem Biophys 428:204\u0026ndash;208. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.abb.2004.06.008\u003c/span\u003e\u003cspan address=\"10.1016/j.abb.2004.06.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee S, Kim Y-Y, Lee Y, An G (2007) Rice P1B-type heavy-metal ATPase, OsHMA9, is a metal efflux protein. Plant Physiol 145:831\u0026ndash;842. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1104/pp.107.102236\u003c/span\u003e\u003cspan address=\"10.1104/pp.107.102236\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLei GJ, Sun L, Sun Y, Zhu XF, Li GX, Zheng SJ (2020) Jasmonic acid alleviates cadmium toxicity in arabidopsis via suppression of cadmium uptake and translocation. J Integr Plant Biol 62:218\u0026ndash;227. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/jipb.12801\u003c/span\u003e\u003cspan address=\"10.1111/jipb.12801\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu X, Gao Y, Zhao X, Zhang X, Ben L, Li Z, Dong G, Zhou J, Huang J, Yao Y (2023) Validation of novel reference genes in different rice plant tissues through mining RNA-seq datasets. Plants 12:3946. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants12233946\u003c/span\u003e\u003cspan address=\"10.3390/plants12233946\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMashabela MD, Masamba P, Kappo AP (2023) Applications of metabolomics for the elucidation of abiotic stress tolerance in plants: a special focus on osmotic stress and heavy metal toxicity. Plants 12:269\u0026ndash;286. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants12020269\u003c/span\u003e\u003cspan address=\"10.3390/plants12020269\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMustafa AM, Abouelenein D, Angeloni S, Maggi F, Navarini L, Sagratini G, Santanatoglia A, Torregiani E, Vittori S, Caprioli G (2022) A New HPLC-MS/MS Method for the Simultaneous Determination of Quercetin and Its Derivatives in Green Coffee Beans. Foods 11:3033. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/foods11193033\u003c/span\u003e\u003cspan address=\"10.3390/foods11193033\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark CH, Yeo HJ, Park YJ, Morgan AMA, Valan Arasu M, Al-Dhabi NA, Park SU (2017) Influence of Indole-3-Acetic Acid and Gibberellic Acid on Phenylpropanoid Accumulation in Common Buckwheat (Fagopyrum esculentum Moench) Sprouts. Molecules 22:374. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/molecules22030374\u003c/span\u003e\u003cspan address=\"10.3390/molecules22030374\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQixing Z, Yanyu W, Xianzhe X (1994) Compound pollution of Cd and Zn and its ecological effect on rice plant. Chin J Appl Ecol 5:438\u0026ndash;441\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamel F, Birtic S, Ginies C, Soubigou-Taconnat L, Triantaphylid\u0026egrave;s C, Havaux M (2012) Carotenoid oxidation products are stress signals that mediate gene responses to singlet oxygen in plants. Proc. Natl. Acad. Sci. 109, 5535\u0026ndash;5540. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.1115982109\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1115982109\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRizwan M, Ali S, Rehman MZ, ur, Maqbool A (2019) A critical review on the effects of zinc at toxic levels of cadmium in plants. Environ Sci Pollut Res 26:6279\u0026ndash;6289. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11356-019-04174-6\u003c/span\u003e\u003cspan address=\"10.1007/s11356-019-04174-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSasaki A, Yamaji N, Ma JF (2014) Overexpression of \u003cem\u003eOsHMA3\u003c/em\u003e enhances Cd tolerance and expression of Zn transporter genes in rice. J Exp Bot 65:6013\u0026ndash;6021. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/jxb/eru340\u003c/span\u003e\u003cspan address=\"10.1093/jxb/eru340\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShahzad M, Bibi A, Khan A, Shahzad A, Xu Z, Maruza TM, Zhang G (2025) Utilization of antagonistic interactions between micronutrients and cadmium (Cd) to alleviate Cd toxicity and accumulation in crops. Plants 14:707\u0026ndash;721. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants14050707\u003c/span\u003e\u003cspan address=\"10.3390/plants14050707\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTavarez M, Grusak MA, Sankaran RP (2023) The effect of exogenous cadmium and zinc applications on cadmium, zinc and essential mineral bioaccessibility in three lines of rice that differ in grain cadmium accumulation. Foods 12:4026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/foods12214026\u003c/span\u003e\u003cspan address=\"10.3390/foods12214026\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTavarez M, Grusak MA, Sankaran RP (2022) Effects of zinc fertilization on grain cadmium accumulation, gene expression, and essential mineral partitioning in rice. Agronomy 12:2182. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/agronomy12092182\u003c/span\u003e\u003cspan address=\"10.3390/agronomy12092182\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTian X, Zhang J, Ye Z, Fang W, Ding X, Yin Y (2025) Zinc sulfate stress enhances flavonoid content and antioxidant capacity from finger millet sprouts for high-quality production. Foods 14:2563. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/foods14152563\u003c/span\u003e\u003cspan address=\"10.3390/foods14152563\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUeno D, Yamaji N, Kono I, Huang CF, Ando T, Yano M, Ma JF (2010) Gene limiting cadmium accumulation in rice. Proc. Natl. Acad. Sci. 107, 16500\u0026ndash;16505. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.1005396107\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1005396107\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUmair Hassan M, Aamer M, Umer Chattha M, Haiying T, Shahzad B, Barbanti L, Nawaz M, Rasheed A, Afzal A, Liu Y, Guoqin H (2020) The critical role of zinc in plants facing the drought stress. Agriculture 10:396\u0026ndash;415. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/agriculture10090396\u003c/span\u003e\u003cspan address=\"10.3390/agriculture10090396\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUraguchi S, Mori S, Kuramata M, Kawasaki A, Arao T, Ishikawa S (2009) Root-to-shoot Cd translocation via the xylem is the major process determining shoot and grain cadmium accumulation in rice. J Exp Bot 60:2677\u0026ndash;2688. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/jxb/erp119\u003c/span\u003e\u003cspan address=\"10.1093/jxb/erp119\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVance TM, Chun OK (2015) Zinc intake is associated with lower cadmium burden in US adults. J Nutr 145:2741\u0026ndash;2748. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3945/jn.115.223099\u003c/span\u003e\u003cspan address=\"10.3945/jn.115.223099\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVu KT, Lan PDT, Nguyen NTH, Thanh HN (2022) Cadmium immobilization in the rice - paddy soil with biochar additive 23. 85\u0026ndash;89. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.12911/22998993/146331\u003c/span\u003e\u003cspan address=\"10.12911/22998993/146331\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang B, Lin L, Yuan X, Zhu Y, Wang Y, Li D, He J, Xiao Y (2023) Low-level cadmium exposure induced hormesis in peppermint young plant by constantly activating antioxidant activity based on physiological and transcriptomic analyses. Front Plant Sci 14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpls.2023.1088285\u003c/span\u003e\u003cspan address=\"10.3389/fpls.2023.1088285\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang H, Liu M, Zhang Y, Jiang Q, Wang Q, Gu Y, Song X, Li Y, Ye Y, Wang F, Chen X, Wang Z (2024) Foliar spraying of Zn/Si affects Cd accumulation in paddy grains by regulating the remobilization and transport of Cd in vegetative organs. Plant Physiol Biochem 207:108351. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.plaphy.2024.108351\u003c/span\u003e\u003cspan address=\"10.1016/j.plaphy.2024.108351\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang H, Xu C, Luo Z-C, Zhu H-H, Wang S, Zhu Q-H, Huang D-Y, Zhang Y-Z, Xiong J, He Y-B (2018) Foliar application of Zn can reduce Cd concentrations in rice (\u003cem\u003eOryza sativa\u003c/em\u003e L.) under field conditions. Environ Sci Pollut Res Int 25:29287\u0026ndash;29294. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11356-018-2938-6\u003c/span\u003e\u003cspan address=\"10.1007/s11356-018-2938-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Q, Zeng X, Song Q, Sun Y, Feng Y, Lai Y (2020) Identification of key genes and modules in response to cadmium stress in different rice varieties and stem nodes by weighted gene co-expression network analysis. Sci Rep 10:9525. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-020-66132-4\u003c/span\u003e\u003cspan address=\"10.1038/s41598-020-66132-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang S, Wu M, Zhong S, Sun J, Mao X, Qiu N, Zhou F (2023) A rapid and quantitative method for determining seed viability using 2,3,5-triphenyl tetrazolium chloride (TTC): with the example of wheat seed. Molecules 28:6828. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/molecules28196828\u003c/span\u003e\u003cspan address=\"10.3390/molecules28196828\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Z, Wang Y, Yang J, Hu K, An B, Deng X, Li Y (2016) Reliable selection and holistic stability evaluation of reference genes for rice under 22 different experimental conditions. Appl Biochem Biotechnol 179:753\u0026ndash;775. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12010-016-2029-4\u003c/span\u003e\u003cspan address=\"10.1007/s12010-016-2029-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWen B, Jiang H, Gao Y, Zhou Q, Qie H (2024) Source analysis and bioavailability of soil cadmium in poyang lake plain of China based on principal component analysis and positive definite matrix factor. Minerals 14:514\u0026ndash;526. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/min14050514\u003c/span\u003e\u003cspan address=\"10.3390/min14050514\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWright MN, Ziegler A (2017) ranger: a fast implementation of random forests for high dimensional data in C\u0026thinsp;+\u0026thinsp;+\u0026thinsp;and. R J Stat Softw 77:1\u0026ndash;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18637/jss.v077.i01\u003c/span\u003e\u003cspan address=\"10.18637/jss.v077.i01\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXue W, Zhang X, Zhang C, Wang C, Huang Y, Liu Z (2023) Mitigating the toxicity of reactive oxygen species induced by cadmium via restoring citrate valve and improving the stability of enzyme structure in rice. Chemosphere 327:138511. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chemosphere.2023.138511\u003c/span\u003e\u003cspan address=\"10.1016/j.chemosphere.2023.138511\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYoshida S, Forno D, Cock J (1971) Laboratory manual for physiological studies of rice, Laboratory manual for physiological studies of rice. Los Ba\u0026ntilde;os, Philippines\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang H, Sun X, Hwarari D, Du X, Wang Y, Xu H, Lv S, Wang T, Yang L, Hou D (2023) Oxidative stress response and metal transport in roots of macleaya cordata exposed to lead and zinc. Plants 12:516. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/plants12030516\u003c/span\u003e\u003cspan address=\"10.3390/plants12030516\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang S, Wu X, Peng J, Meng X, Shi B, Zhou L, Bai L (2021) Study of the physiological dynamics of cadmium accumulation in two varieties of rice with different cadmium-accumulating properties. J. Chem. 2021, 6238893. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1155/2021/6238893\u003c/span\u003e\u003cspan address=\"10.1155/2021/6238893\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Rice, Cadmium accumulation, Zinc–cadium interaction, Genotype-dependent response, Random Forest, Sparse partial least squares","lastPublishedDoi":"10.21203/rs.3.rs-9361934/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9361934/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCadmium (Cd) contamination poses a major risk to rice safety, while zinc (Zn) can modify Cd accumulation in a genotype-dependent manner, with its biological basis remaining incompletely understood. Here, 44 rice varieties were hydroponically cultured under Cd stress either alone or under a near-equimolar Zn\u0026thinsp;+\u0026thinsp;Cd co-exposure treatment, and Zn\u0026ndash;Cd interaction patterns were analyzed using an integrated framework combining response-landscape quantification, Random Forest modeling, and sparse partial least squares (sPLS) analysis. The response landscape revealed a four-quadrant distribution of interaction types across cultivars, with response intensities spanning nearly a 200-fold range, indicating that Zn effects on Cd accumulation were highly variable rather than uniformly inhibitory. Random Forest classification distinguished response direction with ~\u0026thinsp;70% accuracy using baseline metal status and Zn-induced physiological shifts as predictors, suggesting that non-random information about response direction is recoverable from surface-level phenotypic variation. sPLS analysis of 12 representative cultivars further resolved layer-specific candidate signal structures, with translocation-related variation associated primarily with a root superoxide/flavonoid-centered structure, leaf Cd accumulation with a more flavonoid/hormone-centered structure, and total Cd amplitude with a phenolic/flavonoid-associated structure. Together, these results show that Zn effects on Cd accumulation are genotype-dependent and identify the root redox baseline state under Cd stress alone as a candidate upstream feature associated with response direction. They further suggest that contrasting outcomes across cultivars may reflect different state-dependent response regimes linked to baseline redox status. This framework provides a basis for understanding why the same Zn intervention can yield opposite Cd outcomes across rice cultivars, and it requires broader validation before any screening or management application is proposed.\u003c/p\u003e","manuscriptTitle":"Genotype-Dependent Dual Effects of Zinc on Cadmium Accumulation in Rice Revealed by a Multi-Scale Quantitative Framework","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-17 18:57:00","doi":"10.21203/rs.3.rs-9361934/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"83d64805-fce3-43ff-b061-8a2663e020e5","owner":[],"postedDate":"April 17th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Rejected","date":"2026-05-01T09:11:58+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-30T11:07:52+00:00","index":20,"fulltext":""},{"type":"reviewerAgreed","content":"316617033512668796964006608451271931051","date":"2026-04-29T00:39:45+00:00","index":19,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-01T09:25:12+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-17 18:57:00","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9361934","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9361934","identity":"rs-9361934","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}